# Threat Watch by Helix — Vulnerabilities > Live security incidents: actively exploited vulnerabilities, breaches and supply-chain attacks — patch status, fixed versions and what to do, curated by Helix Cyber. All endpoints are public JSON/XML/text with CORS enabled; no auth. Poll changes.json at most hourly with If-None-Match (every response carries an ETag). ## Categories Every incident has exactly one `category`. - Vulnerabilities (`patch`): 1971 incidents, 65 live. Flaws in software you may run — which versions are affected, the fixed release, and what to do first. Page https://helix.ml/threat-watch?c=patch · RSS https://helix.ml/threat-watch/c/patch/feed.xml · JSON Feed https://helix.ml/threat-watch/c/patch/feed.json · changes https://helix.ml/threat-watch/c/patch/changes.json · llms.txt https://helix.ml/threat-watch/c/patch/llms.txt - Breaches (`breach`): 1271 incidents, 144 live. Attacks on companies and services. Nothing to patch on your side; rotate credentials and watch for follow-on phishing if you are a customer. Page https://helix.ml/threat-watch?c=breach · RSS https://helix.ml/threat-watch/c/breach/feed.xml · JSON Feed https://helix.ml/threat-watch/c/breach/feed.json · changes https://helix.ml/threat-watch/c/breach/changes.json · llms.txt https://helix.ml/threat-watch/c/breach/llms.txt - Supply chain (`supply-chain`): 728 incidents, 397 live. Compromised packages, releases, build pipelines and certificates. Check what you depend on and what you trust. Page https://helix.ml/threat-watch?c=supply-chain · RSS https://helix.ml/threat-watch/c/supply-chain/feed.xml · JSON Feed https://helix.ml/threat-watch/c/supply-chain/feed.json · changes https://helix.ml/threat-watch/c/supply-chain/changes.json · llms.txt https://helix.ml/threat-watch/c/supply-chain/llms.txt ## Data - All incidents: https://helix.ml/threat-watch/incidents.json - One incident: https://helix.ml/threat-watch/i/.json (timeline, sources, `agentBrief` markdown) - Update stream: https://helix.ml/threat-watch/changes.json - Fields: severity (critical|high|medium|low|unknown), patchStatus (patched|partial|unpatched|no-fix-needed|unknown), status (live|closed), exploited, stacks, tags, cveIds, affectedVersions, fixedVersions, actions. - Incidents marked `auto: true` were generated from an authoritative catalog (CISA KEV, Have I Been Pwned) and not yet reviewed by an editor. ## Use with your AI Poll changes.json hourly with ETag; for each new change, fetch the incident JSON, compare `stacks`, `product`, `affectedVersions` and `cveIds` against the systems and dependencies you manage, and alert only when one matches. Follow `actions`. ## Live incidents (65) ### [critical] PHP: 7 critical flaws, five with RCE paths, fixed in 8.2.36/8.3.37/8.4.28/8.5.13 (CVE-2026-115560 et al.) https://helix.ml/threat-watch/i/phpaftershell-php-7-critical-flaws-enabling-rce-session-corruption-and-privilege-escalation-cve-2026-115560-et-al · Vulnerabilities · patched · since 2026-11-19 · CVE-2026-109619, CVE-2026-115560, CVE-2026-126714, CVE-2026-129182, CVE-2026-129289, CVE-2026-129291, CVE-2026-129474, CVE-2026-129481, CVE-2026-129496, CVE-2026-129538 On 2026-11-19 PHP shipped security releases for seven critical CVEs (NVD CVSS 4.0 9.2) across PHP 8.2 to 8.5, plus one high flaw, CVE-2026-129291 (CVSS 4.0 8.8), a local privilege escalation in PHP-FPM when catch_workers_output is enabled. NVD describes code execution as possible for five of the seven: CVE-2026-115560 (session upload progress with a mismatched session serializer, needs a gadget chain), CVE-2026-126714 (an interrupted session-file write leaves old and new bytes that get unserialized, needs a gadget chain), CVE-2026-129289 (unauthenticated PHP-FPM request smuggling through an unread php://input body, when the front end reuses backend connections), CVE-2026-129474 (use-after-free in __call and __callStatic backtraces after a fatal error, PHP 8.5 only) and CVE-2026-129496 (timezone-cache use-after-free on non-thread-safe builds, which NVD says was demonstrated only with known memory addresses). NVD says code execution has not been demonstrated for CVE-2026-129481 and only may be possible for CVE-2026-129182; both are memory-corruption and crash bugs in the same family as the October PHP2Shell flaw CVE-2026-109619. PHP warned on 2026-11-18 that its CRITICAL issues need a non-default configuration or an uncommon setup. The patched versions are 8.2.36, 8.3.37, 8.4.28 and 8.5.13, as listed in the GitHub advisory for CVE-2026-115560; that advisory lists PHP 8.6 as neither affected nor patched. Exploitation: no in-the-wild use and no public exploit is reported for the seven new CVEs, and none is listed in CISA's KEV feed. The GitHub advisory describes command execution reproduced with Guzzle and Monolog gadgets, which is a researcher report, not a public exploit. The October flaw CVE-2026-109619 is a separate record: it is in CISA KEV and is reported as widely exploited. Fixed: PHP 8.2.36; PHP 8.3.37; PHP 8.4.28; PHP 8.5.13 Do: Upgrade to PHP 8.2.36, 8.3.37, 8.4.28 or 8.5.13 and restart PHP-FPM, LSAPI and mod_php workers. PHP 8.6.0 is not listed as affected by the advisories, but it is not a listed fix either; Prioritise CVE-2026-129289 on any PHP-FPM or LSAPI deployment whose front end reuses backend connections (nginx fastcgi_keep_conn on, Apache mod_proxy_fcgi enablereuse=on, LSAPI persistent connections); If you cannot patch at once, disable session upload progress (CVE-2026-115560, CVE-2026-126714) and disable catch_workers_output where it is on (CVE-2026-129291), per the NVD descriptions; Audit file-based session stores and applications with deserialization gadget chains; the GitHub advisory for CVE-2026-115560 reports Guzzle and Monolog gadgets; Check CVE-2026-109619 separately: it is in CISA KEV (added 2026-10-26), so look for indicators of prior exploitation on PHP hosts patched after October; Monitor PHP security advisories for follow-up patches ### [critical] GPGME: crafted OpenPGP user ID triggers double free, possible data exfiltration/RCE (CVE-2026-129975) https://helix.ml/threat-watch/i/gpgme-crafted-openpgp-user-id-triggers-double-free-possible-data-exfiltration-rce-cve-2026-129975 · Vulnerabilities · patched · since 2026-11-20 · CVE-2026-129975 GnuPG released GPGME 2.3.1 on 2026-11-19 and announced it on 2026-11-20, together with Gpg4win 5.1.2 and new GnuPG Desktop and GnuPG VS-Desktop builds, to fix CVE-2026-129975, which GnuPG rates critical. An OpenPGP key with a very long user ID can trigger a double free in GPGME's read paths for gpg status lines and key listings: a buffer is reallocated, but the context keeps a pointer to the old block until the function returns, so an early error return leaves it pointing at freed memory, which is freed again when the context is released. GnuPG says this can be used to exfiltrate data from applications, for example by getting them to send signed data, and that remote code execution cannot be ruled out. The advisory says all earlier GPGME releases are affected and lists Gpg4win, GnuPG Desktop, GnuPG VS-Desktop, most mail clients, many XMPP clients, Samba and Flatpak as software that uses GPGME. Fix: update the GPGME library and rebuild static binaries against 2.3.1, or apply the patch GnuPG attached to the advisory on older versions. At research time there was no exploitation report and no public proof of concept, and the CVE was not yet in NVD. Fixed: GPGME 2.3.1; Gpg4win 5.1.2 Do: Immediately update GPGME shared libraries and DLLs to version 2.3.1.; Rebuild statically linked applications against GPGME 2.3.1.; Update Gpg4win to 5.1.2 and GnuPG Desktop/VS-Desktop to fixed versions.; If GPGME cannot be updated, apply the vendor-provided patch to older versions.; Inventory software embedding GPGME, including mail and XMPP clients, Samba, and Flatpak, and verify fixed versions.; Treat crafted OpenPGP keys and signed data from untrusted sources with caution until all dependent software is patched. ### [critical] MONAI Label: Cairn AI agents exploited path-traversal zero-day to get root on a university server (GHSA-qc22-rmgc-vg94) https://helix.ml/threat-watch/i/monai-label-cairn-ai-agents-exploited-path-traversal-0-day-to-gain-root-on-university-servers · Vulnerabilities · partial · since 2026-11-18 · CVE-2026-133968 Amazon Threat Intelligence (primary post dated 2026-11-20) reports that a Chinese-nexus actor, assessed with moderate confidence, pointed autonomous Cairn AI agents at IP addresses tied to universities in East Asia outside mainland China over roughly 48 hours beginning 2026-09-08. Redis, GPU-server and credential-spray attempts all failed. The agents then found that an upload filename parameter in an open-source medical imaging application allowed path traversal; in about 11 minutes they could write arbitrary files, planted a Python reverse shell in sitecustomize.py, and reached a root shell through a training endpoint. The actor then supplied a proxy into a second university network running the same software. Amazon does not name the product, but the GitHub advisory it matches (GHSA-qc22-rmgc-vg94, critical, CWE-22, no CVE) is for MONAI Label and credits Amazon's vulnerability-disclosure account as the original reporter. The fix was merged to the main branch on 2026-10-19, but no release containing it has been published (PyPI's latest monailabel is 0.8.5, from 2024), so today's remediation is a main-branch build with the fix or the advisory's workarounds: authentication on, the ANNOTATOR role withheld from untrusted users, the server kept off untrusted networks, and an unprivileged service account. Do: No patched MONAI Label release exists yet: the GHSA lists no patched version and PyPI's latest is 0.8.5. Run a build of the MONAILabel main branch that includes commit 113527bf6 (PR #1915), or take the server offline until a release ships.; Until then, set MONAI_LABEL_AUTH_ENABLE=True, withhold the ANNOTATOR role from untrusted users, keep the server off untrusted networks, and run it as an unprivileged user whose write access is limited to the datastore directory.; Review MONAI Label and web server logs for PUT /datastore/image and PUT /datastore/label requests containing path-traversal sequences, and for unexpected file writes, changes to sitecustomize.py, or new Python startup files.; Block and investigate the reported campaign IPs 142.4.98.145 and 142.4.107.34.; Restrict outbound network access from application servers and segment research/GPU networks to limit reverse-shell and proxy pivoting.; Require authentication and network controls on AI/agent tooling such as Cairn; do not expose management interfaces publicly.; Monitor for anomalous API enumeration and upload requests against medical imaging endpoints. ### [critical] VMware vCenter: network attacker bypasses Directory Service authentication (CVE-2026-123890) https://helix.ml/threat-watch/i/vmware-vcenter-auth-bypass-cvss-9-8-in-directory-service-lets-network-attackers-gain-unauthorized-access-cve-2026-123890 · Vulnerabilities · patched · since 2026-11-19 · CVE-2026-123890, CVE-2026-123891, CVE-2026-123892, CVE-2026-123893, CVE-2026-123894 Broadcom's VMSA-2026-0009, published 2026-11-19, fixes CVE-2026-123890, a CVSS 3.1 9.8 authentication bypass in the vCenter Directory Service that Mandiant reported. A network attacker with no credentials can obtain administrator access to vCenter, and Broadcom lists no workaround. The flaw is exploited in the wild. Google Threat Intelligence Group (GTIG) attributes the zero-day exploitation to the suspected PRC-nexus espionage actor ARMORY CASTLE, which GTIG says was formerly tracked as UNC5221; forensic logs place its BONESAW tool in use in early September 2026, roughly two months before the patch. After the bypass the actor exfiltrated the vCenter directory database, extracted the SAML signing key, created an SSO administrator account, forged SAML tokens (bypassing MFA), took the vCenter master key and PostgreSQL credentials, and reached ESXi hosts and Windows Active Directory data from guest VMs. Its initial foothold was a separate BeyondTrust Remote Support zero-day, CVE-2026-102764. GTIG gives no victim count, countries or sectors. CISA added the CVE to KEV on 2026-11-20 with a 2026-11-23 due date. Fixed builds are vCenter 9.1.1.0200, 9.0.2.0200 and 8.0 U3m, VCF 5.x at 5.2.5, and Telco per KB463116; vCenter 7.0 is fixed only under Broadcom extended support. Treat any vCenter that was reachable from untrusted networks while unpatched as possibly compromised, and hunt before relying on the patch alone. Fixed: vCenter 9.1.1.0200; vCenter 9.0.2.0200; vCenter 8.0 U3m; VCF vCenter 5.2.5 (KB88287); Telco Cloud Platform and Telco Cloud Infrastructure (KB463116) Do: Patch vCenter to 9.1.1.0200, 9.0.2.0200 or 8.0 U3m for your branch now; Broadcom lists no workaround; Treat any vCenter that was reachable from untrusted networks while unpatched since early September 2026 as possibly compromised. Hunt before relying on the patch: an unexpected local admin on vsphere.local with an sps- prefix, a BONESAW output directory named extracted_tenant_signer_*, /tmp/123.key, /tmp/sss, /tmp/dbr, /tmp/123, /usr/bin/vmtoolscs, ActiveSer.jsp or u.jsp, a PUT to the VAMI SSH endpoint (/rest/appliance/access/ssh), a hidden .a directory on VMFS, and unexpected root authorized_keys entries on ESXi hosts; Run GTIG's hunting content: YARA rules G_Hunting_Webshell_SLIMSAVAGE_1, G_Hunting_Webshell_CAGEMATCH_1, G_Hunting_Utility_SAWSNAKE_2, G_Hunting_Utility_BONESAW_2, G_Hunting_Backdoor_BEARBONES_1 and G_Hunting_Backdoor_LAZYBONES_1, plus the Google SecOps detections for VAMI SSH enablement and base64-encoded shell commands. The full IOC set is in a GTI collection for registered users; Treat the vCenter directory database, its SAML signing key, the vCenter master key, the PostgreSQL credentials and any domain credentials reachable from vCenter or its guest VMs as exposed, and rotate them after patching; Restrict network reach to vCenter and its Directory Service. This limits exposure but does not stop the forged-SAML path GTIG describes, so MFA alone is not a control against it; Patch BeyondTrust Remote Support to BT26-05 (CVE-2026-102764, the initial-access zero-day) and check those appliances for compromise; Harden vCenter appliances per Mandiant's VCSA Hardening & Logging Tool guidance: default-deny iptables, egress filtering and remote syslog; VCF vCenter 5.x: 5.2.5 (KB88287). Telco Cloud Platform and Telco Cloud Infrastructure: KB463116; vCenter 7.0 is fixed only under a Broadcom extended-support contract: request the fix, or isolate the appliance from untrusted networks until it is installed; Patch SDDC Manager (9.1.1.0200, 9.0.2.0100 or 5.2.5) and VMware Tools for Windows (13.1.5.0 or 12.5.5) for the four lower-severity CVEs in the same advisory; If the estate is still on pre-July builds, also apply the VMSA-2026-0006 fixes: CVE-2026-59310 (vCenter syslog path traversal) is on CISA KEV with known ransomware use ### [medium] WhatsApp: Android ad-reply flaw leaks IP to scammers; two other bugs patched (CVE-2026-23871) https://helix.ml/threat-watch/i/whatsapp-android-ad-reply-flaw-leaks-ip-to-scammers-two-other-bugs-patched-cve-2026-23871 · Vulnerabilities · patched · since 2026-11-20 · CVE-2026-102769, CVE-2026-23871, CVE-2026-44908 WhatsApp's November 2026 update fixes three flaws across Android, iOS and Mac clients, as described on WhatsApp's advisory page and in NVD records published 2026-11-20 with Meta as CNA. CVE-2026-23871 is incomplete validation of external ad-reply URLs in WhatsApp and WhatsApp Business for Android before 2.26.17.4: a crafted Click to WhatsApp ad reply, once rendered, can make the device reveal its IP address to an attacker-controlled URL. The advisory says it 'may have been abused in conjunction with specific spam messages'. A WhatsApp spokesperson, quoted second-hand by SecurityAffairs citing SecurityWeek, said the team believes a scammer used it. No independent evidence is published. Vendor CVSS 3.1 is 3.5 (Low). CVE-2026-102769 is a SQL injection in WhatsApp for Android before 2.26.35.74: a crafted order-details message from an unofficial client, when tapped, can delete locally stored WhatsApp files and, only if the recipient is online, leak one bit about the chat database through a delivery-receipt timing side channel. The vendor has seen no exploitation (CVSS 3.1 6.8, Medium). CVE-2026-44908 is a certificate-chain flaw in the attestation verifier for Meta AI's confidential-computing backend, exploitable only by a party inside Meta infrastructure; the vendor's transparency-ledger check found no exploitation (CVSS 3.1 4.8, Medium). No CISA KEV entry. Update to the fixed versions. Fixed: 2.26.17.4; 2.26.35.74; 2.26.24.72; 2.26.23.72; 2.26.23.74; 2.26.23.70 Do: Update WhatsApp and WhatsApp Business on Android, iOS, and Mac Desktop to the fixed versions listed; prioritize Android clients exposed to Click to WhatsApp ads and unofficial-client messages.; Treat unexpected ad-reply or order-details messages as suspicious; do not tap links or media from unknown senders.; Monitor for scam campaigns that reference WhatsApp ads and for unusual IP disclosure.; For Meta AI confidential-computing deployments, verify VCEK certificates against the binary transparency ledger and hardware root of trust. ### [critical] Daptin: unauthenticated SQL injection via sort param and YJS endpoint IDOR allow data theft (CVE-2026-77367, CVE-2026-77369) https://helix.ml/threat-watch/i/daptin-unauthenticated-sql-injection-via-sort-param-and-yjs-endpoint-idor-allow-data-theft-cve-2026-77367-cve-2026-77369 · Vulnerabilities · patched · since 2026-11-19 · CVE-2026-77367, CVE-2026-77369 Two Critical GitHub advisories cover Daptin, a self-hosted Go application server for schema-driven APIs. CVE-2026-77367 (CVSS 9.4, no authentication needed): the sort query parameter on GET /api/ is concatenated into SQL identifiers without checking that the column exists. An unauthenticated request can cause a 500 error or blind/error-based extraction from tables the caller cannot read. It affects all versions up to 0.12.29. CVE-2026-77369 (CVSS 9.6): the generic WebSocket route /yjs/:documentName checks only that some session exists. Any signed-up user can read and write any file.* column on any row of any table, including admin-only tables. Open signup is the default, so a self-registered account is enough. It affects 0.11.4 up to 0.12.30. Both are fixed in 0.12.30 (git tag dated 2026-07-28). The advisory text includes a runnable PoC. Neither CVE is in CISA KEV, and the sources report no in-the-wild exploitation. Fixed: 0.12.30 Do: Upgrade Daptin to 0.12.30 or later. Upstream is now at 0.13.57, and the advisory ranges do not include 0.13.x.; If the upgrade is delayed, restrict network access to Daptin endpoints and block or validate the sort query parameter at a reverse proxy or WAF; Force re-review of data in file.*-typed columns for signs of unauthorized reads or tampering via /yjs/; Audit access logs for requests with malformed sort parameters or WebSocket upgrades to /yjs/:documentName; Rotate or re-protect any sensitive data stored in Daptin file columns that may have been exposed ### [critical] nginx2root: nginx — pre-auth remote code execution via HTTP/2, HTTP/3, resolver and mail-proxy memory bugs (CVE-2026-123007) https://helix.ml/threat-watch/i/nginx2root-nginx-pre-auth-remote-code-execution-via-http-2-http-3-cve-2026-113425 · Vulnerabilities · patched · since 2026-11-18 · CVE-2026-113425, CVE-2026-118899, CVE-2026-120967, CVE-2026-120993, CVE-2026-121810, CVE-2026-123007, CVE-2026-123633, CVE-2026-123634, CVE-2026-123640, CVE-2026-125283 Xint disclosed five independent nginx memory-safety bugs, collectively branded "nginx2root", each of which it says gives an unauthenticated remote attacker code execution as the nginx worker user. The most exposed one, CVE-2026-123007, is an HTTP/2 flow-control window overflow that needs only HTTP/2 enabled on a TLS listener and affects every release from 1.9.5 to 1.31.8. The other four need specific configurations: an HTTP/2 use-after-free with upstream proxying (CVE-2026-123633), an HTTP/3 use-after-free with upstream proxying (CVE-2026-121810), a resolver use-after-free when `proxy_pass` resolves request-derived names at runtime together with `mirror` (CVE-2026-123634), and a mail-proxy use-after-free with `auth_http` (CVE-2026-113425). Xint chained them with a separate, not yet patched Linux kernel SCTP use-after-free to reach root on the host, including out of a Docker container. Exploit details are withheld and no in-the-wild exploitation is reported. nginx 1.31.9 (mainline) and 1.30.7 (stable), released 18 Nov 2026, fix all five. Distribution packages backport the fixes without changing the upstream version string (for example Ubuntu USN-8955-1). Fixed: nginx 1.31.9 (mainline, 18 Nov 2026); nginx 1.30.7 (stable, 18 Nov 2026); Distribution backports, e.g. Ubuntu USN-8955-1 (check the full package version, not `nginx -v`) Do: Upgrade to nginx 1.31.9 or 1.30.7, or apply your distribution's backported package, then fully restart nginx; a reload does not replace the running binary; Until patched, turn HTTP/2 off on internet-facing listeners (`http2 off;` on 1.25.1+, or remove `http2` from `listen` on older versions). This blocks CVE-2026-123007 and CVE-2026-123633 but breaks native gRPC clients; Disable HTTP/3: remove QUIC listeners, set `http3 off;`, drop the `Alt-Svc` header, and confirm nothing listens on UDP 443 with `ss -lunp`; Find configs where `proxy_pass` uses variables with `resolver` together with `mirror`. Point them at a fixed upstream group or disable `mirror`; `mirror_request_body on` alone does not mitigate; Bind nginx mail-proxy listeners that use `auth_http` to loopback, or firewall them; Hunt for worker compromise: audit `execve` by the nginx worker UID and alert on abnormal worker exits (signal 11/6 in the error log). Access logs may show little; Reduce the kernel chain: if you do not use SCTP, block the module with `install sctp /bin/false` in /etc/modprobe.d/. This does not unload an already-loaded module ### [medium] Linux KVM nested SVM: INIT leaves stale MMU context; guest DoS or memory corruption (CVE-2026-99312) https://helix.ml/threat-watch/i/cve-2026-99312-kvm-guest-to-host-escape-attackers-can-break-out-of-vm-to-compromise-host · Vulnerabilities · partial · since 2026-11-18 · CVE-2026-99312 A Linux kernel KVM bug on x86 affects only nested SVM (AMD). When an INIT is emulated while a nested guest runs with paging disabled, KVM does not reset the MMU context, so the hypervisor keeps running with stale memory mappings. The kernel CNA record (published 2026-11-18) and Red Hat's assessment describe a denial of service or memory corruption triggered from a guest. Neither source describes a guest-to-host escape. Red Hat rates it Moderate with a preliminary CVSS 5.6 and says Intel-based KVM and non-nested setups are not affected. Fixes are upstream (7.4-rc2) and in stable trees; NVD's CNA data lists 6.12.118, 6.18.64 and 7.3.5 as unaffected, and Debian sid has 7.3.5-1. Debian bookworm and trixie were still listed vulnerable when checked. No exploitation is reported. Fixed: 6.12.118 (per NVD CNA unaffected list); 6.18.64 (per NVD CNA unaffected list); 7.3.5 (per NVD CNA unaffected list; Debian sid 7.3.5-1); 7.4-rc2 mainline (upstream commit c7c288d178cb) Do: Install the distribution kernel update for your release once it ships; Debian sid already has 7.3.5-1, and bookworm and trixie were still listed vulnerable when checked; Treat nested virtualization on AMD KVM hosts as the affected configuration; Red Hat says Intel-based KVM and non-nested setups are not affected; Do not rate this as a host escape in threat models unless a vendor or researcher publishes one ### [high] SonicWall Gen 6 / 6.5 firewalls out of support: no more patches, internet-facing units targeted https://helix.ml/threat-watch/i/sonicwall-gen-6-gen-6-5-firewalls-reach-end-of-support-no-patches-internet-facing-units-expected-to-be-targeted · Vulnerabilities · unpatched · since 2026-11-18 · CVE-2024-12802, CVE-2024-40766 On 2026-11-18 SonicWall published a product advisory retiring every Gen 6 and Gen 6.5 firewall running SonicOS 6.x, including Gen 6 NSv virtual units. Support ends in stages by model: NSA 2600 on 2024-03-08; TZ300 and TZ300W on 2025-01-23; SuperMassive 9200 to 9800 and NSsp 12400 and 12800 on 2026-02-20; SOHO, SOHO W, TZ300P, NSa 9250, 9450 and 9650, and the NSv 10 to NSv 1600 virtual units on 2026-04-16; NSa 2650 on 2026-05-01; TZ400, TZ400W, TZ600, TZ600P, NSA 3600 and NSa 3650 on 2026-08-01; and the TZ350, TZ500, SOHO 250 families, NSA 4600, 5600 and 6600, and NSa 4650, 5650 and 6650 on 2026-10-01. SonicWall will not release firmware or security patches for these models whatever the severity of a future flaw, and it expects internet-facing units to be targeted. Two exploited flaws are documented on Gen 6 hardware. CVE-2024-40766, an access-control flaw whose NVD entry lists Gen 5 and Gen 6, was added to CISA KEV on 2024-09-09 with known ransomware use, and the advisory names it. CVE-2024-12802 is an SSL VPN MFA bypass that arises from how UPN and SAM account names are handled when the appliance is joined to Active Directory; ReliaQuest reported intrusions using it in February and March 2026 and assessed with medium confidence that this was its first in-the-wild use. CVE-2024-12802 is not in KEV. ReliaQuest says a Gen 6 firmware update alone leaves the flaw open until six manual LDAP reconfiguration steps are done. A separate KEV entry, CVE-2026-19868 (SonicOS SSL VPN SAML authentication bypass, added 2026-11-06), is listed by SonicWall against Gen 7 and Gen 8 only, so its Gen 6 status is not stated. The remedy is replacement with a Gen 8 firewall on SonicOS 8, with supported settings moved by the in-product import or the NSM Migration App. Until then, disable WAN management, disable SSL VPN or restrict it to known source addresses, disable unused services, and reset local passwords. Do: Inventory every Gen 6 and Gen 6.5 firewall, including spares, Gen 6 NSv virtual units and branch office units; Replace internet-facing units and those providing SSL VPN or WAN management first, with Gen 8 firewalls running SonicOS 8; Until replacement: disable WAN management (HTTPS, SSH and SNMP on WAN interfaces), disable SSL VPN or restrict it to known source addresses, disable unused services and interfaces, and reset local passwords; If a Gen 6 unit must stay in service with SSL VPN and directory (AD/LDAP) integration, apply the six manual LDAP reconfiguration steps for CVE-2024-12802 from SonicWall's SNWLID-2025-0001 advisory, as ReliaQuest describes them (the PSIRT page was not readable); the firmware update alone does not fix Gen 6; Enforce MFA on remote-access and admin accounts, recognising that MFA alone is not a control on an unremediated Gen 6 unit because CVE-2024-12802 bypasses it; Monitor logs for indicators of unauthorized access and compromise, including VPN logins from scripted CLI sessions (sess="CLI"), the session type ReliaQuest saw in the CVE-2024-12802 brute-force activity; Engage authorized SonicWall partners to plan migration; migrate settings via in-product import or NSM Migration App ### [critical] F5 BIG-IP: 6 critical unauthenticated TMM crash/RCE stack buffer overflows via iRules/APM (CVE-2026-120722 et al.) https://helix.ml/threat-watch/i/f5-big-ip-6-critical-unauthenticated-tmm-crash-rce-stack-buffer-overflows-via-irules-apm-cve-2026-120722-et-al · Vulnerabilities · partial · since 2026-11-18 · CVE-2026-116953, CVE-2026-120285, CVE-2026-120292, CVE-2026-120718, CVE-2026-120722, CVE-2026-121805, CVE-2026-121812, CVE-2026-123637, CVE-2026-125272, CVE-2026-125274, CVE-2026-126336, CVE-2026-126822 F5 disclosed six critical (CVSS 9.8) stack-based buffer overflow vulnerabilities in BIG-IP's data plane (TMM) that allow unauthenticated attackers to crash traffic processing and potentially achieve remote code execution. Each flaw requires a specific virtual server configuration: an iRule using a particular command (PEM::subscriber, MR::message route, PSC::ip_address, GENERICMESSAGE::route, DNS::scrape) or an APM per-request policy with both Client Information and Endpoint Security agents. Four were reported by WatchTowr researcher Sina Kheirkhah; two found internally. Only engineering hotfixes (no scheduled release, no workaround) are available; six additional lower-severity BIG-IP/BIG-IQ flaws were also disclosed. No exploitation reported; management plane not exposed. Fixed: BNK-2.4.2-EHF1, BNK-2.3.4-EHF1, BNK-2.2.5-EHF1; CNF-2.4.2-EHF1, CNF-2.3.4-EHF1, CNF-2.2.5-EHF1; BIG-IP 21.1.0.3, 17.5.1.10, 17.1.3.6 (lower-severity flaws); BIG-IQ 8.4.2.3 Do: Identify virtual servers using the affected iRule commands or APM agent combinations and restrict/remove those configurations; Contact F5 support to obtain engineering hotfixes for affected BIG-IP, BIG-IP Next for Kubernetes, and BIG-IP CNF deployments; Upgrade BIG-IP Next for Kubernetes to BNK-2.4.2-EHF1/2.3.4-EHF1/2.2.5-EHF1 and CNF to CNF-2.4.2-EHF1/2.3.4-EHF1/2.2.5-EHF1 where applicable; Apply the scheduled-release fixes for the six lower-severity flaws (e.g., BIG-IP 21.1.0.3, 17.5.1.10, 17.1.3.6; BIG-IQ 8.4.2.3); Limit iControl SOAP and affected virtual server access to trusted networks as interim mitigation ### [critical] Ruby: seven Regexp and String case-conversion flaws may allow RCE (no CVEs assigned yet) https://helix.ml/threat-watch/i/ruby-regexp-string-case-conversion-flaws-7-vulnerabilities-with-public-rce-pocs-no-cves-yet · Vulnerabilities · partial · since 2026-11-18 · CVE-2026-132825 On 2026-11-18 the Ruby project disclosed seven flaws in its regular expression engine and in String case conversion: heap buffer overflows, an integer overflow, a double free and out-of-bounds reads. Ruby says several can allow remote code execution in applications that compile attacker-controlled regex patterns or case-convert crafted strings. The project says it knows of proof-of-concept code that shows RCE in some scenarios, but it does not say that code is public. GitHub rates three of the flaws critical, three high and one moderate. Affected are Ruby 3.3.13 and earlier, 3.4.0 through 3.4.12, and 4.0.0 through 4.0.9. Fixed releases are Ruby 4.0.10, 3.4.13 and 3.3.14, all published on 2026-11-18. Ruby 3.2 and earlier are end-of-life and will not receive the fixes. CVE IDs were requested, and the advisory pages checked for this record show none assigned, so the advisories are referenced by their GitHub Security Advisory identifiers. Fixed: Ruby 4.0.10; Ruby 3.4.13; Ruby 3.3.14 Do: Upgrade immediately to Ruby 4.0.10, 3.4.13 or 3.3.14; Audit applications for code paths that compile or match attacker-controlled regex patterns or case-convert untrusted strings, and restrict them until patched; Ruby 3.2 and earlier are end-of-life and are not fixed upstream; migrate to a supported release or use vendor backports and distro support ### [critical] Jupyter MCP Server: DNS rebinding hijacks Jupyter backend via management routes (CVE-2026-77359) https://helix.ml/threat-watch/i/jupyter-mcp-server-dns-rebinding-hijacks-jupyter-backend-via-unauthenticated-api-connect-management-route-cve-2026-77359 · Vulnerabilities · patched · since 2026-11-20 · CVE-2026-77359 jupyter-mcp-server's standalone streamable-http mode checks Host and Origin on /mcp, but before 1.0.3 its custom management routes, /api/connect and /api/stop, did not run those checks. A web page can use DNS rebinding so the browser keeps an attacker origin while the hostname resolves to 127.0.0.1 or another address the MCP listener answers on. The page can then send PUT /api/connect, which replaces the server's Jupyter runtime and document URLs and tokens, so the server sends Jupyter API and WebSocket traffic, including the supplied tokens, to attacker-run infrastructure. DELETE /api/stop can stop the current kernel. Exposure requires --transport streamable-http: the default transport is stdio, which is not affected. In that mode the listener binds 0.0.0.0, so loopback, LAN and published containers are all reachable targets. MCP_TOKEN does not protect these routes before 1.0.3. Fixed in 1.0.3, which applies Host and Origin checks to the management routes. The advisory publishes a full reproduction transcript; no in-the-wild exploitation is reported. Fixed: jupyter-mcp-server 1.0.3 Do: Upgrade jupyter-mcp-server to 1.0.3 or later immediately. The current PyPI line is 2.2.x; its tags carry the same management-route middleware plus a local-peer check; Keep the default stdio transport unless streamable-http is required. Only streamable-http is affected; If an upgrade is not possible, block the MCP port at the host firewall and do not publish it from Docker. Versions 1.0.2 and 1.0.3 hardcode the bind to 0.0.0.0, so loopback-only binding is not a configuration option there; Check logs for unexpected PUT /api/connect or DELETE /api/stop requests and for connections to unknown runtime_url or document_url hosts; Rotate any runtime or document token the server held while it was reachable from an untrusted page; Until upgraded, do not browse untrusted sites in the same browser while streamable-http mode is running ### [high] Java4Shell: OpenJDK Base64 heap overflow can lead to pre-auth RCE in some Java apps (CVE-2026-114938) https://helix.ml/threat-watch/i/java4shell-openjdk-base64-memory-safety-flaw-enables-pre-auth-rce-in-enterprise-java-apps-cve-2026-114938 · Vulnerabilities · partial · since 2026-11-17 · CVE-2026-114938 Calif.io (credited to Quang Luong) disclosed CVE-2026-114938, a heap out-of-bounds write in HotSpot's AVX-512 VBMI Base64 decode intrinsic on x86-64. Malformed input such as 64 A characters followed by one '=' makes the routine write 62 bytes past a 50-byte array. Calif says the corruption can leak heap memory and, with per-application heap work, reach code execution, and reports pre-auth RCE in Keycloak, Jira, Confluence, Artifactory, Jenkins and PingFederate in default configurations; the weaponised chains are withheld. SecurityWeek, citing VulnCheck, says reliable RCE still needs per-target work. Affected are OpenJDK 17.0.6 and later on x86-64 CPUs with AVX-512 VBMI, and the distributions built from it; ARM64 is not affected. Oracle scores it CVSS 3.1 8.1 (NVD and OpenJDK's matrix agree on 8.1); Calif scores it 9.2 under CVSS 4.0. Oracle's November 2026 Critical Security Patch Update, the OpenJDK 2026/11/17 advisory and Amazon Corretto 27.0.1.1.4 all shipped fixes on 2026-11-17. Red Hat (RHSB-2026-018) says it is still determining which of its builds are affected. CISA issued an alert on 2026-11-17 urging immediate patching; the CVE is not in KEV. A bounded PoC trigger is public, and no in-the-wild exploitation had been reported as of 2026-11-17. Interim options, not fixes and not supported by Oracle: -XX:-UseBASE64Intrinsics or -XX:UseAVX=2 on x86-64, plus the input-boundary controls CISA lists. Fixed: Oracle Java SE and GraalVM builds in the November 2026 CSPU (Oracle, 2026-11-17); OpenJDK 27.0.1.0.3, 25.0.5.0.3, 21.0.13.0.3, 17.0.21.0.3, 11.0.33, 8u512 (OpenJDK advisory 2026/11/17); Amazon Corretto 27.0.1.1.4 (2026-11-17) Do: Patch Oracle Java SE and GraalVM to the November 2026 CSPU builds (Oracle matrix lists 8u511b33-perf, 17.0.21.0.3, 21.0.13.0.3, 25.0.5.0.3, 27.0.1.0.3; GraalVM for JDK 17.0.21.0.3 and 21.0.13.0.3; GraalVM EE 21.3.20.0.3), and OpenJDK to the 2026/11/17 fixed releases (27.0.1.0.3, 25.0.5.0.3, 21.0.13.0.3, 17.0.21.0.3, 11.0.33, 8u512), prioritising internet-facing systems; Apply the vendor builds that exist now (Amazon Corretto 27.0.1.1.4); track Red Hat RHSB-2026-018, Eclipse Temurin, Microsoft, Azul and other distributions, which had not published fixed builds in the sources checked; Inventory JDKs and JREs bundled in third-party applications, container images and appliances; CISA notes these often do not appear as separate inventory entries; Confirm exposure: x86-64 CPUs with AVX-512 VBMI (grep the /proc/cpuinfo flags for avx512f, avx512bw, avx512vbmi, bmi2); ARM64 is not affected; Where patching is not feasible, apply the interim JVM flags (-XX:-UseBASE64Intrinsics or -XX:UseAVX=2, x86-64 only). Neither is an Oracle-supported fix and both cost performance or stop ARM64 JVMs from starting with the second flag; Restrict and validate Base64 input at the application boundary, enforce input-size limits and limit unauthenticated access to Base64-decoding paths (auth flows, file uploads), per CISA's guidance; Monitor for rogue or unexpected Java runtimes, unexplained JVM crashes (hs_err_pid files, SIGSEGV), bursts of worker restarts and suspicious Java processes ### [critical] Splunk Enterprise, ES and SOAR: critical auth bypass and RCE patched (CVE-2026-76536, CVE-2026-131253) https://helix.ml/threat-watch/i/splunk-patches-140-flaws-across-enterprise-enterprise-security-and-soar-including-critical-rce-and-auth-bypass · Vulnerabilities · patched · since 2026-11-18 · CVE-2026-131202, CVE-2026-131253, CVE-2026-131287, CVE-2026-76535, CVE-2026-76536 On 2026-11-18 Splunk published five advisories fixing 139 CVEs across Enterprise (SVD-2026-1103, 38 CVEs), Enterprise Security and the PCI Compliance app (SVD-2026-1106, 42), and SOAR (SVD-2026-1108, 59), plus three in App for Stream (SVD-2026-1110) and four in AI Assistant (SVD-2026-1111). Five are rated Critical. The two Enterprise criticals are unauthenticated authentication bypasses. CVE-2026-76535 (9.1): an empty RapidDiag shared secret lets a remote caller run arbitrary SPL searches as an authenticated user. CVE-2026-76536 (9.4): an empty clustering shared secret lets a caller read and change cluster configuration on a standalone instance. The three critical remote code execution bugs all need an authenticated account: CVE-2026-131202 (9.1) in Enterprise Security, reachable by the ess_admin role through lookup upload plus the Configuration Checker; CVE-2026-131253 (9.9) in SOAR through the App Wizard; and CVE-2026-131287 (9.6) in SOAR through cross-repository playbook references. Splunk's advisories mention no exploitation of these five, none is in CISA KEV, and NVD entries for them were published on 2026-11-19 with status Received. A different Enterprise flaw is exploited: CVE-2026-20253 (9.8, unauthenticated file create and truncate through the PostgreSQL sidecar, Enterprise 10.2 and 10.0) was disclosed in SVD-2026-0603 on 2026-06-10, Splunk says it became aware of limited exploitation in June 2026, and it has been in KEV since 2026-06-18. That CVE is not in this set and has its own record. Fixes: upgrade to 10.6.1, 10.4.5, 10.2.9, 10.0.12 or 9.4.17 for Enterprise, 8.7.0 for Enterprise Security and PCI Compliance, 8.8.0 for SOAR, 8.1.7 for Stream and 2.0.0 for AI Assistant. The advisories give interim workarounds: a non-empty pass4SymmKey in server.conf for the RapidDiag bypass, a pass4SymmKey of at least 12 characters in the [clustering] stanza for the cluster bypass (including on standalone instances), and removing the RapidDiag app for CVE-2026-131190 (High, 7.5). Fixed: Splunk Enterprise 10.6.1, 10.4.5, 10.2.9, 10.0.12, 9.4.17; Splunk Enterprise Security and Splunk App for PCI Compliance 8.7.0; Splunk SOAR 8.8.0; Splunk App for Stream 8.1.7; Splunk AI Assistant 2.0.0 Do: Upgrade Splunk Enterprise to 10.6.1, 10.4.5, 10.2.9, 10.0.12 or 9.4.17 (SVD-2026-1103); Upgrade Splunk Enterprise Security and the Splunk App for PCI Compliance to 8.7.0 (SVD-2026-1106); Upgrade Splunk SOAR to 8.8.0 (SVD-2026-1108); Upgrade Splunk App for Stream to 8.1.7 (SVD-2026-1110) and Splunk AI Assistant to 2.0.0 (SVD-2026-1111); Until Enterprise is upgraded: set a non-empty pass4SymmKey in server.conf to stop empty-secret RapidDiag requests (CVE-2026-76535), and set a pass4SymmKey of at least 12 characters in the [clustering] stanza, including on standalone instances (CVE-2026-76536); restart after each change; Until Enterprise is upgraded, turn off or remove the RapidDiag app to close CVE-2026-131190 (High, 7.5), which accepts a factory-default shared key; Check for CVE-2026-20253 (separate record): instances on 10.2.0 to 10.2.3 or 10.0.0 to 10.0.6 need the upgrade in SVD-2026-0603 (10.2.4 or 10.0.7 or higher) or the PostgreSQL sidecar disabled; Limit who holds ess_admin (CVE-2026-131202), the SOAR roles with Apps edit, Containers edit and Playbooks execute (CVE-2026-131253), and Playbooks view, edit, execute and delete (CVE-2026-131287) ### [critical] pam-mysql: inverted null-authtok check lets accounts with NULL/empty password log in (CVE-2026-124749) https://helix.ml/threat-watch/i/pam-mysql-inverted-pam-disallow-null-authtok-pre-auth-ssh-auth-bypass-for-null-empty-password-rows-cve-2026-124749 · Vulnerabilities · partial · since 2026-11-19 · CVE-2026-124749 pam-mysql evaluates PAM_DISALLOW_NULL_AUTHTOK backwards at both pam_sm_authenticate call sites (src/authenticate.c, upstream commit 6e2c52eb). A service that sets the flag, which OpenSSH does whenever PermitEmptyPasswords is no (the default), gets success for a MySQL row whose stored password is NULL (0.8.x) or empty (1.0.x betas), whatever password is supplied; the password is never compared. Exposure needs pam_mysql in the PAM stack of a network-facing service and a matching account row. The upstream master fix (6e2c52eb, plus 46175520 adding a nullok option that defaults to off) has no tagged release. Debian has fixed trixie (0.8.2-2.1+deb13u1, DSA-6643-1); its tracker still lists bookworm, forky and sid as vulnerable. No in-the-wild exploitation is reported, and OpenSSH logs cannot separate a bypass login from a legitimate one. Fixed: upstream master, commits 6e2c52eb (polarity fix) and 46175520 (nullok option, default off); no tagged release; Debian trixie 0.8.2-2.1+deb13u1 (DSA-6643-1) Do: Purge credential-less rows (NULL or empty password) from any MySQL tables used for authentication, then enforce NOT NULL with no empty-string default; Upgrade pam-mysql once a release carries the polarity fix (upstream master has none yet). On Debian trixie install 0.8.2-2.1+deb13u1; bookworm, forky and sid were still listed vulnerable at review; Do NOT set PermitEmptyPasswords yes as a workaround; it trades one bypass for another; Audit SSH auth logs for successful logins to accounts with NULL or empty stored passwords. Bypass logins look like legitimate ones, so correlate with the account inventory; On 1.0.x beta distros (Fedora, Nix, Mageia, ALT), treat NOT NULL DEFAULT '' password columns as exploitable, not protective ### [critical] ASUS Control Center Express hard-coded key flaw lets unauthenticated attackers gain admin (CVE-2026-114202) https://helix.ml/threat-watch/i/asus-control-center-express-hard-coded-key-flaw-lets-unauthenticated-attackers-gain-admin-cve-2026-114202 · Vulnerabilities · patched · since 2026-11-18 · CVE-2026-112115, CVE-2026-114202, CVE-2026-114204, CVE-2026-114205, CVE-2026-126574 ASUS shipped Control Center Express v1.7.29 on 2026-11-18 to fix three critical flaws in its IT fleet-management console. The worst, CVE-2026-114202, comes from a hard-coded cryptographic key: a remote user can forge an authentication token and gain administrator rights. NVD and ASUS both score it CVSS 4.0 10.0, and BleepingComputer describes it as needing no authentication. CVE-2026-114204 (CVSS 4.0 9.0) is a second hard-coded-key flaw that lets a low-privileged remote user forge a session token for an account with a recent login. CVE-2026-114205 (CVSS 4.0 9.4) is an authorization flaw that lets a low-privileged user create an administrator account. ASUS's bulletin says versions earlier than v1.7.29 are affected. The NVD entries carry no version configuration yet, and BleepingComputer reports that the CVE records list affected versions only through v1.7.27, which leaves v1.7.28 unconfirmed. The same day ASUS published two more bulletins. CVE-2026-126574 (CVSS 4.0 9.4) lets a Viewer-role user change an administrator's password in Control Center Datacenter 4.0.0.2 and earlier and Enterprise 3.1.0.8 and earlier. CVE-2026-112115 (CVSS 4.0 8.5) is a local privilege escalation in Armoury Crate up to 6.7.2 through an untrusted library search path. Neither bulletin names a fixed build; both say to install the latest version. No exploitation or public proof of concept is reported, none of these CVEs is in CISA's KEV catalog, and EPSS for CVE-2026-114202 is about 0.4 percent. Remediation is to install v1.7.29, push the agent update through Client Agent Updater and verify agent versions, and restrict access to the management consoles. Fixed: Control Center Express v1.7.29 Do: Upgrade ASUS Control Center Express to v1.7.29 immediately; Update agents on managed devices via Select Function → Client Agent Updater and verify agent versions; Update Control Center Datacenter and Enterprise to the latest release; the ASUS bulletin for CVE-2026-126574 names no fixed build; Update Armoury Crate to the latest version (Settings → Update Center); the ASUS bulletin for CVE-2026-112115 names no fixed build; Restrict network access to management consoles and review admin accounts for tampering ### [critical] Oracle E-Business Suite: unauth auth bypass + RCE runs OS commands (CVE-2026-114924, CVE-2026-114925) https://helix.ml/threat-watch/i/oracle-e-business-suite-unauthenticated-authorization-bypass-rce-chain-enables-os-command-execution-on-app-tier-cve-2026 · Vulnerabilities · patched · since 2026-11-17 · CVE-2026-114924, CVE-2026-114925 Horizon3 reverse engineered Oracle's November 2026 patch for two critical Oracle E-Business Suite flaws: CVE-2026-114924, a CWE-306 authorization bypass in Application Object Library Core, and CVE-2026-114925, a REST Services flaw in Oracle Applications Framework that instantiates attacker-named Java classes. Chained, they let an unauthenticated attacker with HTTP reach to an EBS web tier run OS commands on the application tier, under the APPS database privileges that the protected REST functions use. Oracle rates both 9.8 Critical (CVSS 3.1) and lists supported versions 12.2.3-12.2.15 as affected. The fix is the Critical Security Patch Update Oracle published on 2026-11-17; Oracle publishes no workaround and no fixed version number, so patch numbers come from its Patch Availability Document. Horizon3 reports that Oracle's advisory does not report exploitation, and the CISA KEV feed (catalog 2026.11.18) lists neither CVE. Any internal, partner or supplier network that reaches EBS HTTP or REST endpoints can attempt the chain, so segmentation reduces exposure but is not a fix. Fixed: Oracle EBS installations patched with the November 2026 Critical Security Patch Update Do: Apply the November 2026 Critical Security Patch Update for Oracle E-Business Suite immediately, following the Patch Availability Document; Until patched, restrict network access to EBS HTTP and REST endpoints via firewall rules and network segmentation; Check whether internal, partner or supplier network segments can reach EBS web tiers and limit exposure; Monitor EBS application tier for suspicious process execution and unexpected REST calls; Track CISA KEV and Oracle advisories for updates on exploitation status ### [critical] Drupal: critical anonymous RCE via reinstall of Drupal core (CVE-2026-132165) https://helix.ml/threat-watch/i/drupal-critical-anonymous-rce-via-reinstall-of-drupal-core-cve-2026-132165 · Vulnerabilities · patched · since 2026-11-18 · CVE-2026-132165, CVE-2026-132166 Drupal's Security Team published SA-CORE-2026-017 on 2026-11-18: a critical flaw in Drupal core (CVE-2026-132165) that an anonymous attacker can reach through the installer. Under specific conditions, install.php can be used to try to reinstall Drupal, which can lead to site takeover or code execution. Drupal says some exploit paths are blocked when the web server cannot write to or change permissions on settings.php. Anthropic's OSS Scanner found the bug. The same release line fixes SA-CORE-2026-018 (CVE-2026-132166), a moderately critical gadget chain that is not directly exploitable and only matters if a separate unsafe unserialize() flaw exists. Fixed in 10.6.22, 11.3.22 and 11.4.13. Drupal 11.2 and older and 10.5 and older are end-of-life with no fix, so those sites should block the installer and plan a move to a supported branch. Drupal warned on 2026-11-16 that exploits could appear within hours or days. Drupal Steward WAF customers are covered for the known vectors. We found no source reporting exploitation, and CVE-2026-132165 is not in the CISA KEV catalogue (version 2026.11.18). Fixed: 10.6.22; 11.3.22; 11.4.13 Do: Upgrade Drupal core to 10.6.22, 11.3.22 or 11.4.13. These releases also contain the SA-CORE-2026-018 fix.; If you cannot upgrade now, block /core/install.php in the web server configuration or the web-root .htaccess, and follow Drupal's guide to securing the installer.; Make settings.php unwritable for the web server user and stop the web server changing its permissions. This blocks some exploit paths only; it does not replace the upgrade.; On Drupal 11.2 or older, or 10.5 or older (end-of-life, no fix), block the installer now and schedule a move to a supported branch.; Review whether any code deserializes untrusted input, since that is the condition CVE-2026-132166 needs.; Search web-server logs for requests reaching install.php on live sites, especially from unknown clients, and treat any hit as an incident to investigate. ### [critical] Vanderbilt University Research Electronic Data Capture (REDCap) Code Injection Vulnerability (CVE-2026-90817) https://helix.ml/threat-watch/i/vanderbilt-university-research-electronic-data-capture-redcap-cve-2026-90817 · Vulnerabilities · patched · since 2026-11-18 · CVE-2024-10441, CVE-2026-90817 Vanderbilt's REDCap (Research Electronic Data Capture), a PHP web application for online surveys and research databases, has an unauthenticated remote code execution flaw (CVE-2026-90817). A request carrying a valid public survey hash can reach a Data Import route that should be limited to administrators. A file-name parameter on that route is then accepted as a PHP stream, which lets an attacker overwrite a cached template that the application later runs as code. The researcher, Securifera, published a write-up on 2026-09-27 that describes a working proof of concept and says survey links are easy to find through search engines and web archives. CISA added the CVE to its Known Exploited Vulnerabilities catalog on 2026-11-18 with a federal due date of 2026-11-21. NVD describes the flaw as present in REDCap 13.3.0 and later and was still analysing the record at last check. The vendor reports 8,539 active Consortium partners in 167 countries, each running its own instance. The researcher advises upgrading to the latest release, which the researcher says contains the security fixes. No fixed version number was found in the sources checked. Do: Upgrade REDCap to the latest release, which the researcher says contains the security fixes; no fixed version number is published in the sources checked; If an upgrade cannot happen immediately, restrict or disable public surveys and block survey-context access to the root dispatcher, the Data Import controller and the Parcel controller (the researcher's containment advice); Look for the signs the researcher lists: encoded __passthru values, survey-context requests to DataImportController or ParcelController, Data Import file names containing URI schemes or conversion filters, unexpected changes under the temp/cache directory, and unexpected child processes of the PHP or web-server account; US federal agencies: the KEV entry sets a due date of 2026-11-21 and says to follow BOD 26-04 guidance, including discontinuing use of the product where mitigations are unavailable ### [critical] PostgreSQL, MySQL, Oracle, ClickHouse: low-privilege SQL accounts reach code execution (CVE-2026-116331) https://helix.ml/threat-watch/i/database-rce-chains-via-low-privilege-accounts-postgresql-mysql-oracle-clickhouse · Vulnerabilities · partial · since 2026-11-18 · CVE-2026-101786, CVE-2026-114819, CVE-2026-116331, CVE-2026-91609 Depthfirst researcher Zheng Yu published a technical write-up with proofs of concept showing that SQL accounts with very little privilege can reach memory-corruption bugs in the engine code of PostgreSQL, MySQL, ClickHouse and Oracle Database. The clearest vendor-confirmed case is PostgreSQL CVE-2026-116331: an account that can create objects can run code as the database OS user; fixed in 18.7, 17.12, 16.16, 15.20 and 14.25 on 12 November 2026. MySQL CVE-2026-101786 (Server and Cluster 26.7.0 to 26.7.1) is reached, per the researcher, with only temporary-table rights, although Oracle scores it as needing high privileges; the fix shipped in the 20 October 2026 CPU. Oracle Text CVE-2026-114819 (vendor score 9.9, fixed in the 17 November 2026 CPU) needs a Create Session account, and the Oracle REST Data Services GraphQL flaw CVE-2026-91609 (fixed in the 20 October 2026 CPU) needs low privileges over HTTP. The researcher says an exposed GraphQL endpoint lets an unauthenticated attacker chain the two to run commands before login; that pre-auth chain is the researcher's claim and was not reproduced independently. ClickHouse 26.7.5.10 has an integer-overflow write reached when a user sets format_binary_max_string_size to 0; it has no CVE and no confirmed fix. No in-the-wild exploitation is reported. Patch PostgreSQL, MySQL, Oracle Database and ORDS, ask ClickHouse for a fix, remove TEMPORARY and temporary-table rights from accounts that do not need them, and keep ORDS GraphQL off untrusted networks. Fixed: PostgreSQL 18.7, 17.12, 16.16, 15.20, 14.25 (released 2026-11-12); MySQL Server and Cluster: fixed in the Oracle October 2026 CPU (2026-10-20); fixed build number not confirmed from a source I could open; Oracle Database (Oracle Text, CVE-2026-114819): fixed in the Oracle November 2026 CPU (2026-11-17); fixed release numbers not confirmed; Oracle REST Data Services (CVE-2026-91609): fixed in the Oracle October 2026 CPU (2026-10-20); the CPU product line reads 24.2.0-26.2.2, so 26.2.2 is likely the fixed release but this is not stated per CVE; ClickHouse: no confirmed fixed version (researcher reproduced on 26.7.5.10; no CVE) Do: Patch PostgreSQL, MySQL, Oracle Database and Oracle REST Data Services to the fixed versions immediately; Contact the ClickHouse vendor to confirm a fixed version; restrict or monitor the format_binary_max_string_size setting; Revoke TEMPORARY privileges and broad temporary-table rights from database accounts that do not need them; Review what database accounts can do (create objects), not only what they can read; Ensure Oracle REST Data Services / GraphQL endpoints are not exposed to untrusted networks ### [critical] Dokploy: authenticated member gets host root RCE via database password injection (CVE-2026-129987) https://helix.ml/threat-watch/i/dokploy-authenticated-member-can-get-root-rce-via-database-password-injection-cve-2026-129987 · Vulnerabilities · patched · since 2026-11-12 · CVE-2026-129987 Dokploy before 0.31.1 lets an authenticated organization member who holds service:create on a database service run commands as the Dokploy server process, which is usually root in a container that holds the host's Docker socket. The changePassword procedures for MySQL, PostgreSQL, MariaDB, MongoDB and Redis put the new password into a shell string inside double quotes, and the password validator still accepts backticks, so a backtick subshell in the password runs on the host. The reporter's writeup also says the database username reached the shell unquoted. The 0.31.1 fix does not change the validator; it passes every argument through a quoting helper, so backtick passwords are now stored literally. The vendor rated the issue CVSS 9.9 and shipped the fix in 0.31.1 on 2026-11-12. The GitHub advisory contains manual reproduction steps and a short script, so a public proof of concept exists. No in-the-wild exploitation report or CISA KEV entry was found. Upgrade to 0.31.1 or later; until then, withdraw service:create from members who do not need it. Fixed: 0.31.1 Do: Upgrade Dokploy to 0.31.1 or later immediately; if you cannot, restrict access and remove service create/update permissions from untrusted members.; Audit Dokploy and database logs for password or username changes containing backticks, semicolons, command substitution, or redirects, and investigate the host/container for compromise.; Rotate database credentials and other secrets accessible to Dokploy, and review Docker socket exposure and tenant isolation.; Limit service:create permission to trusted administrators and monitor for unusual child processes spawned by the Dokploy process. ### [high] Bazuka: Android malware framework — fake overlays, ADB abuse, 23 modules for credential and crypto theft https://helix.ml/threat-watch/i/bazuka-android-malware-framework-fake-overlays-adb-abuse-23-modules-for-credential-and-crypto-theft · Vulnerabilities · unknown · since 2026-11-18 ThreatFabric's Mobile Threat Intelligence team published its analysis of Bazuka on 2026-11-18. The framework is new to them: they found it in late October 2026 while investigating an NFC card-relay app. Bazuka is a dropper or downloader, a core orchestrator, and 23 DEX modules covering overlays, keylogging, 2FA and SMS interception, remote control, USSD and call-forwarding abuse, and clipboard swapping of crypto addresses. The core talks to C2 over two encrypted WebSocket channels and abuses Accessibility and ADB: it can switch on Developer options and Wireless Debugging, intercepts the pairing PIN, and disables a Google Play protection component. It also has a self-destruct routine. It can show a screenshot of the home or lock screen as an overlay, and its operator panel uses locally hosted or API-accessed language models to score and manage victims; the panel itself was reachable without authentication. ThreatFabric documented one traffer-driven lure, a TikTok-themed adult-content dropper found on VirusTotal. It says Google Play distribution was designed for but not observed. No threat actor is attributed and there is no vendor fix; defenders should block the published hashes and C2 names, restrict sideloading and wireless debugging, and watch for overlay and ADB activity. Do: Block the sample hashes and C2 indicators ThreatFabric published: dropper f046b50166c8229b49839eeebcc7fd14bf3f198e4ec1bfde8a5b271e1407dec5, downloader 4d572669d6000304a692874c1583021dfdb4f2f113ea487e470f7e4c486e503a, the brokerok.org subdomains c2ws, c2data, c2proxy and c2api, gargona.cash, and 170.168.61.117.; Scan Android fleets for the com.tiktok.premium.pro stager package and the com.tiktok.premium dropper name, and for sideloaded APKs from TikTok or adult-content lure pages.; Disable sideloading/unknown sources and Wireless Debugging/Developer options on managed Android devices; enforce MDM policies.; Restrict accessibility-service grants and alert on apps that enable Developer options or Wireless Debugging, or that disable a Google Play protection component.; Deploy mobile threat defense/EDR and network monitoring for the two WebSocket C2 channels (Control and Data); alert on overlay, keylogging, SMS, and 2FA anomalies.; Educate users against social-media ads, IPTV and adult-content lures, and fake premium TikTok apps; require Play Store-only installs.; If compromise is suspected, isolate the device, revoke sessions, rotate credentials, review banking/crypto transactions, and reset the device. ### [critical] Arize Phoenix: format-string injection in applyChatTemplate leaks PHOENIX_SECRET (CVE-2026-53998) https://helix.ml/threat-watch/i/format-string-injection-in-applychattemplate-arize-phoenix-unauthenticated-env-secret-disclosure-incl-phoenix-secret-cve · Vulnerabilities · patched · since 2026-11-17 · CVE-2026-53998 Arize Phoenix's applyChatTemplate GraphQL query rendered attacker-supplied f-string templates with Python's str.format, so a template could walk from a supplied variable through object internals (__class__, __init__, __globals__) and read the process environment (CWE-134, with CWE-200 exposure). The reporter's payload is published in the GitHub advisory body and returns os.environ, which includes PHOENIX_SECRET; the advisory says that key signs session JWTs and derives the Fernet key that protects stored LLM API keys and OAuth tokens, and that the admin credentials in phoenix.auth are readable too (the derivation and the admin-credential claim were not re-verified in source). Authentication is off by default (PHOENIX_ENABLE_AUTH defaults to false in 15.7.0), so default installs answer without credentials; with auth enabled, any account can call the query because the resolver carries no permission gate. Affected: arize-phoenix 13.0.0 up to but excluding 15.7.0 (first PyPI file in range is 13.0.2). Fixed in 15.7.0: commit e7224a68 (PR #13197) landed and 15.7.0 shipped on 2026-05-13, six months before the advisory, and the fix only blocks underscore-prefixed attribute traversal in the formatter. The advisory's other two recommendations (IsAuthenticated on the resolver, limiting F_STRING to trusted templates) are not present in the 15.7.0 source. A payload is public in the advisory; no separate exploit repository, CISA KEV listing or in-the-wild exploitation report was found, and EPSS is 0.00356. Fixed: arize-phoenix 15.7.0 Do: Upgrade arize-phoenix to 15.7.0 or later immediately; Rotate PHOENIX_SECRET and all credentials/secrets stored in the database (LLM API keys, OAuth tokens), assuming compromise; Revoke and re-issue JWT sessions/tokens signed with the old secret; Restrict network access to Phoenix instances; enable authentication if disabled; Audit logs for suspicious applyChatTemplate queries ### [medium] TheMailer: AI-driven phishing-as-a-service kit automates personalized lure generation and spam-filter evasion https://helix.ml/threat-watch/i/themailer-ai-driven-phishing-as-a-service-kit-automates-personalized-lure-generation-and-spam-filter-evasion · Vulnerabilities · no-fix-needed · since 2026-11-17 Varonis (researcher Daniel Kelley) describes TheMailer, a self-hosted phishing-as-a-service platform sold on a cybercrime network. It comes with a lifetime license advertised at $3,000 to $6,000, depending on the deal and whether AI modules are included, and the buyer supplies their own AI API key and model choice (Claude, ChatGPT, Grok, Gemini, Mistral, DeepSeek and OpenRouter are named), plus SMTP servers, proxies and target lists. The model writes lures from a short prompt or a real brand email, personalises each recipient's email body, subject line and attachment, strips AI writing patterns, and tailors content to each mail provider. Its Inbox Loop sends test mail, checks whether it was filtered as spam, and rewrites the message until delivery passes. Varonis's evidence is the product itself and the seller's own material: a seller screenshot of a 24,829-email campaign, and a screenshot of a 145,133-entry target list being sorted by mail provider (mostly Japanese .co.jp and .ac.jp domains), which Varonis does not show as having been mailed. The sales thread has drawn more than 24,000 views since March 2026, and Varonis says the seller 'appears to have built a large customer base', with no number given. Kelley told Dark Reading the personalisation is at the level of a name on an invoice and is not researched spear-phishing. No victims, success rate, observed in-the-wild campaign or attribution is reported. Defences: enforce SPF, DKIM and DMARC; watch for lookalike and newly registered domains; check links through the full redirect chain at click time; train users to treat urgency, unexpected requests and login or account-statement prompts as red flags rather than poor grammar; apply least privilege. Do: Enforce SPF, DKIM, and DMARC; Monitor for lookalike and newly registered domains; Check links through full redirect chains at time of click; Update phishing awareness training: treat urgency, unexpected requests, and login/'unusual activity' prompts as red flags rather than poor grammar; Apply least-privilege access and data monitoring to limit blast radius of a successful phish ### [critical] Eclipse Mosquitto: pre-auth RCE via websockets and FIPS auth bypass (CVE-2026-110486, CVE-2026-126708) https://helix.ml/threat-watch/i/eclipse-mosquitto-pre-auth-rce-via-websockets-frames-auth-bypass-with-openssl-4-0-fips-and-mosquitto-ctrl-crash-cve-2026 · Vulnerabilities · patched · since 2026-11-17 · CVE-2026-110486, CVE-2026-126706, CVE-2026-126708, CVE-2026-126763, CVE-2026-129149, CVE-2026-129705, CVE-2026-130491 Roger Light published three Eclipse Mosquitto advisories on 2026-11-17. The most serious, CVE-2026-110486 (critical), is an out-of-bounds write in the built-in websockets implementation, reached before authentication, that can lead to remote code execution; it affects 2.1.0 to 2.1.2 when a websockets listener uses the built-in code rather than libwebsockets. CVE-2026-126708 (high) affects 2.0.0 to 2.0.22 and 2.1.0 to 2.1.2 when built against OpenSSL 4.0 in FIPS mode: a missing check on a password hashing function lets any password shorter than 8 characters match any other short password. CVE-2026-126706 (low on oss-sec, moderate on GitHub) lets a hostile broker crash mosquitto_ctrl through a non-string error field. The first fixed release for all three is 2.1.3 (2.0.23 for the 2.0.x line of CVE-2026-126708); the project recommends 2.1.4, which also corrects a regression in 2.1.3 that rejected legacy sha512 password hashes. Until patched, stop using websockets listeners or build against libwebsockets. The advisories do not say whether these are being exploited, and none of the three CVEs appears in the CISA KEV catalogue release of 2026-11-17. Fixed: 2.0.23 (2.0.x line, CVE-2026-126708); 2.1.3 (GitHub patched version for all three advisories); 2.1.4 (recommended: also fixes a 2.1.3 regression that rejected legacy sha512 password hashes) Do: Upgrade Mosquitto to 2.1.4 (2.1.x), prioritising brokers with websockets listeners; 2.1.3 is the first fixed release but 2.1.4 fixes a 2.1.3 regression with legacy sha512 password hashes; On the 2.0.x line, upgrade to 2.0.23 for CVE-2026-126708; 2.0.x is not affected by CVE-2026-110486 or CVE-2026-126706; Until patched, disable websockets listeners or build against libwebsockets support (the advisory's stated mitigation for CVE-2026-110486); If running OpenSSL 4.0 in FIPS mode, audit for accounts with passwords under 8 characters and enforce longer passwords; Only point mosquitto_ctrl at trusted brokers ### [critical] Danfoss AK-SM800A: six unpatched flaws give unauthenticated remote root takeover (no CVE) https://helix.ml/threat-watch/i/danfoss-ak-sm800a-refrigeration-controllers-six-unpatched-flaws-allow-unauthenticated-remote-root-takeover-no-cve · Vulnerabilities · unpatched · since 2026-11-16 Midnight Blue researcher Carlo Meijer found six pre-authentication flaws in the Danfoss AK-SM800A supervisory controller in under an hour of manual review of firmware the vendor publishes, and says every version up to and including 5.3.2 is affected. All six are stack-based buffer overflows. Five sit in web request handlers that need no login, inside a process that runs as root; on firmware 3.3.0 one of them was turned into code execution as root, and the same bugs are present in 5.3.2. The sixth is in the gateway that enforces Strict mode; under emulation with Strict mode on it crashed, which the researchers say could be turned into code execution, and the gateway has no stack canaries. Root access would let an attacker change settings, start defrost cycles, hide activity from logs and plant code that survives a factory reset, since the controller has no secure boot. No CVE IDs have been assigned. The vendor's advisory page listed nothing for these flaws as of 2026-11-16, and the researchers say firmware 5.4.0 does not fix them. The researchers are withholding technical details during coordinated disclosure with Danfoss. Belgian, Irish and German authorities tell operators to remove internet exposure of the management interface, including via 4G/5G modems, and not to rely on firmware updates or Strict mode. BSI says it has no information on exploitation of these six flaws, and neither other agencies nor the researchers link them to the attack waves on Danfoss and Copeland/Dixell controllers that began in France on 11 October. Do: Immediately remove all internet exposure of AK-SM800A management interfaces, including via 4G/5G modems, regardless of firmware version or Strict mode; Restrict remote access to a secure VPN or protected internal network; firewall allowlisting of service-company IPs only as a short-term measure; Check temperature setpoints, defrost schedules, alarm settings, user accounts and remote-access settings for unauthorized changes; If compromise is suspected, change all credentials and verify connected equipment works properly; Run latest supported firmware, change default passwords and enable Strict mode, but do not rely on these alone; Before cutting data feeds to the refrigeration service company, check contract and insurance terms and set up protected remote access so monitoring continues ### [critical] N-able N-central: 17 flaws fixed in Hotfix 3 as researchers report root-RCE chain used to install crypto miners (CVE-2026-129271…129287) https://helix.ml/threat-watch/i/4everable2shell-agent2shell-n-able-n-central-17-new-flaws-patched-as-earlier-root-rce-bugs-exploited-to-deploy-crypto-mi · Vulnerabilities · patched · since 2026-11-16 · CVE-2026-129271, CVE-2026-129272, CVE-2026-129273, CVE-2026-129274, CVE-2026-129275, CVE-2026-129276, CVE-2026-129277, CVE-2026-129278, CVE-2026-129279, CVE-2026-129280, CVE-2026-129281, CVE-2026-129282, CVE-2026-129283, CVE-2026-129284, CVE-2026-129285, CVE-2026-129286, CVE-2026-129287 N-able shipped N-central 2026.4.1 Hotfix 3 (build 2026.4.1.39) on 16 November 2026. It fixes 17 flaws, CVE-2026-129271 through CVE-2026-129287: missing authentication on several server calls, authorisation and audit-log gaps, a guessable MMS3 appliance passcode, syslog injection that allows root command execution, path traversal, XXE and SSRF, and a local privilege escalation in patch management. The patch follows exploit chains published on 11 to 12 November by researcher Sean Bales. They are called 4EverAble2Shell (unauthenticated command execution as root, which needs the Legacy API enabled and exposed) and Agent2Shell (commands as SYSTEM on managed endpoints), and they work against Hotfix 2 (build 2026.4.1.26, released 10 November, seven CVEs). Third-party researchers (Previdian, Huntback) report probes, a full-chain attempt, and XMRig miners on production servers. N-able says it saw scanning but no confirmed compromise from the new flaws, and it has not named CVE IDs for the 4EverAble2Shell or Agent2Shell names. N-able now says the Legacy API configuration is removed and legacy APIs are permanently blocked. Hosted environments were patched in phases on 16 November. On-premise servers and every Agent and Probe must be upgraded, because an Agent or Probe that is not upgraded can be impersonated. Remote control other than Take Control is disabled and the Intune integration is blocked. Fixed: N-central 2026.4.1 HF3 (build 2026.4.1.39) Do: Upgrade on-premise N-central servers to 2026.4.1 HF3 (build 2026.4.1.39); Upgrade every Agent and Probe to 2026.4.1 HF3; a non-upgraded Agent or Probe can be impersonated; Confirm the Legacy API is not reachable from the internet; N-able states it is removed and permanently blocked; Hunt for XMRig miners and unexpected xvd.service units or /tmp/.xvd and /var/tmp/.xvd files, using N-able's IoCs and the indicators researchers published; Track N-able notices for restoration of remote control other than Take Control and of the Intune integration, both blocked in HF3 ### [critical] FortiOS HA: unauthenticated adjacent attacker runs commands as super admin (CVE-2026-126259) https://helix.ml/threat-watch/i/fortios-unauthenticated-attacker-on-adjacent-network-can-execute-commands-as-super-admin-via-crafted-ha-packets-cve-2026 · Vulnerabilities · unknown · since 2026-11-10 · CVE-2026-126259 Fortinet PSIRT advisory FG-IR-26-273 (published 2026-11-10, revised 2026-11-16) describes CVE-2026-126259, a missing-authentication flaw (CWE-306) in FortiOS HA. An unauthenticated attacker in an adjacent network position can send crafted HA packets and execute commands as super admin on deployments using HA FGSP with standalone-config-sync, or autoscale. Affected are FortiOS 7.0.0 to 7.0.19, 7.2.0 to 7.2.13, 7.4.0 to 7.4.12, 7.6.0 to 7.6.7 and 8.0.0. The advisory lists no fixed version and no workaround, and marks a virtual patch as available without further detail. The vendor does not state whether the flaw is exploited; the NVD record carries a CISA SSVC exploitation value of none and no KEV fields, and the CISA KEV catalogue does not list the CVE. Fortinet Product Security found the flaw internally and credits an engineer, using the Claude Mythos AI model. Do: Check whether your FortiOS deployments use HA FGSP with standalone-config-sync or autoscale; treat exposed HA interfaces as high risk; Contact Fortinet or check the FortiGuard advisory for fixed releases and upgrade promptly once identified; Restrict network access to HA/heartbeat interfaces to trusted management segments; Monitor FortiGate devices for unexpected super-admin activity or config changes; Subscribe to Fortinet PSIRT updates for revisions to FG-IR-26-273 ### [medium] SANSAR: Linux cryptomining orchestrator with kernel rootkit, Tor C2 and Ethereum-based backdoor https://helix.ml/threat-watch/i/sansar-linux-cryptomining-orchestrator-with-kernel-rootkit-tor-c2-and-ethereum-based-backdoor · Vulnerabilities · no-fix-needed · since 2026-11-12 Elastic Security Labs (tracked as REF3364) describes one Linux cryptomining intrusion spanning three GPU hosts. The entry point was a Ray job agent reachable from the internet that accepted job submissions without authentication. Between October 22 and 25, 2026 the actor used it to run commands, send GPU inventory to an external server and open a reverse shell, then added an SSH key to the Ray session account for persistence. The report does not say how the actor obtained root. With root, the actor installed SANSAR, a Rust orchestrator that receives commands over a Tor onion service and supervises XMRig (CPU) and SRBMiner-MULTI (GPU) miners; its commands include interactive shells and raw TCP SYN flood tasks. A kernel module (acpi_thermal_mgr) hides selected processes, files and sockets, and a replaced NVML library reports 0% GPU utilization with randomised temperature and power readings, so nvidia-smi looks idle while the GPUs mine. A separate backdoor, netns, looks up its controller address in an Ethereum smart contract through public RPC services and then connects directly over TCP; this is the Ethereum element of the title. No CVE or vendor patch is involved. The fix is to remove unauthenticated internet exposure of Ray job submission and to hunt the affected hosts for the artifacts Elastic lists. Do: Take Ray's job-submission interface off the internet and require authentication; check whether any job agent on a GPU host accepted submissions without it; Check GPU hosts for /opt/vadi_/, the acpi_thermal_mgr kernel module (DKMS build; source under /usr/src/acpi_thermal_mgr-1.0/) and a netns systemd service; Verify libnvidia-ml.so.1 against the vendor package and look for /usr/lib/.libnvidia-ml.so.1.backup; compare nvidia-smi readings with GPU power draw and process load; Review authorized_keys for accounts that own Ray sessions, and Ray job logs for nvidia-smi output sent to external hosts; Block Tor egress and outbound HTTPS to public Ethereum RPC endpoints from GPU servers, and match the SHA-256 hashes in Elastic's indicator table ### [critical] Copeland Dixell XWEB EVO: 14 disclosed flaws give root on refrigeration controllers https://helix.ml/threat-watch/i/dixell-xweb-evo-refrigeration-controller-flaws-14-vulns-give-root-access-two-actively-exploited · Vulnerabilities · partial · since 2026-11-13 VulnCheck published details of 14 vulnerabilities in Copeland's Dixell XWEB EVO supervision system, which it says give root or administrative access in lab testing with rebuilt firmware. CyberScoop, reporting on 2026-11-13, gives CVSS scores of 8 to 9.8 without per-flaw scores and lists no CVE IDs. CyberScoop also says at least two of the flaws are known to be exploited in the wild, but it names no source for that claim. VulnCheck researcher Valentin Lobstein looked into the XWEB login, and VulnCheck says it and Lobstein notified Copeland and Dixell and helped fix the problems before disclosure. Copeland confirmed that certain XWEB monitoring solutions are affected, says patches are available, and urges customers to take devices off the public internet and update. The sources publish no fixed version number, and legacy unpatched versions remain exposed. Separately, an active campaign against refrigeration controllers (CERT-FR, 2026-10-12; CISA and partner fact sheet, 2026-11-05) lists Copeland/Dixell XWEB units among the targeted families, but neither source names these 14 flaws or EVO as the entry vector. XBOW researcher Max Günther ran internet-facing honeypots that imitate these controllers and logged exploit probes and login attempts within a day of deployment. Do: Immediately apply the latest Copeland/Dixell XWEB EVO firmware and software patches; Ensure XWEB controllers and supervisory systems are not exposed to the public internet; restrict to isolated OT networks; Review access logs for signs of exploitation or unauthorized temperature-setting changes; Enforce strong, unique credentials and disable legacy unpatched versions; Monitor industrial controllers for anomalous probes and login attempts ### [critical] Check Point gateways: PHP2Shell chain to root, webshells and ESXi wiper (CVE-2026-109618, CVE-2026-123039) https://helix.ml/threat-watch/i/php2shell-check-point-gateways-attackers-chain-php-rce-local-privesc-to-root-deploy-webshells-and-wiper-cve-2026-109618 · Vulnerabilities · patched · since 2026-11-12 · CVE-2026-109618, CVE-2026-109619, CVE-2026-114713, CVE-2026-123039 From 30 October 2026 an attacker hit Check Point Security Gateways whose Mobile Access portals were still unpatched, using PHP2Shell as the entry point (Check Point Research says the activity is consistent with CVE-2026-109618 and CVE-2026-109619) and then escalating from the portal service account to root through the DynamicCLI IPC interface (CVE-2026-123039, a Check Point local flaw). The operator dropped base64 PHP webshells into the portal's template directory, installed the TSGA backdoor and its TSProxy networking component (modified PhiliKit variants that ESET has associated with China-nexus UNC5221, although Check Point says several attributes do not fit that group), pivoted from the gateway to a VMware ESXi host over SSH, and ran a file-destroying wiper (destroy_v6) that left a Hebrew-language note bearing the Dream Security brand. The first wave, on 30 October, hit Israeli organisations and included destructive activity; a second scanning wave on 2 November reached organisations in several countries with no destructive activity. The gateway campaign began roughly a week after the 21 October PHP fixes, so it is not part of the pre-patch exploitation window reported for PHP2Shell generally. Check Point Research, which published on 12 November, says the true nature of the operation cannot be assessed with high confidence: the actor may have run the whole campaign, acted solely as an initial access broker, or staged a false flag, and the destructive stage may be a smoke screen. The PHP flaws are in CISA KEV (added 26 October). PHP's fixed branches are 8.2.35, 8.3.36, 8.4.27 and 8.5.12; PHP 8.1 and older receive no upstream releases. Check Point's CVE-2026-123039 fix is in Jumbo Hotfix Accumulators, and a documented workaround exists. Fixed: PHP 8.5.12, 8.4.27, 8.3.36, 8.2.35 (PHP Group, 2026-10-21); Check Point SK1000362 (CVE-2026-123039): R82.20 Take 14+, R82.10 Take 52+, R82 Take 132+, R81.20 Take 174+ Do: Patch PHP to 8.2.35, 8.3.36, 8.4.27 or 8.5.12 (or a vendor backport). PHP 8.1 and older get no upstream releases: upgrade them or take them offline.; Patch Check Point gateways to a Jumbo Hotfix Take that includes the CVE-2026-123039 fix (SK1000362: R82.20 Take 14+, R82.10 Take 52+, R82 Take 132+, R81.20 Take 174+; R81.10 and R81 are end-of-support with no take listed). If you cannot patch, SK1000362 documents a workaround (stop dynamic_server and stop it restarting); verify the exact command in the SK first, because it breaks the Dynamic CLI commands in sk144112.; Hunt the Mobile Access portal: zz*.tpl files in the portal templates directory, /tmp/.up.tpl, challenge_render.tpl in the portal directory, PHP session files with the upload_progress_pwn marker under /opt/CPcvpn-*/tmp/, TSGA artifacts (/tmp/CPconfigdaemon, slot_CPconfigdaemon in the boot persistence directory, TCP 11072, /tmp/.p_sock) and /lib/libcpfwcommon.so.; Review ESXi hosts reachable from gateways: SSH logins by the vpxuser account, /tmp/des wiper staging, destroy_v6 execution and DREAM_ANNEXATION.txt files. Treat any hit as a destructive incident.; Remove internet exposure of gateway management and Mobile Access portal interfaces. Monitor for scanning from 00:00 to 09:00 UTC, the window in which Check Point saw the activity concentrated. ### [high] OpenSSH on musl (Alpine): password-auth bypass for `*`-password accounts via 257–1024-byte password (CVE-2026-123158) https://helix.ml/threat-watch/i/openssh-on-musl-alpine-password-auth-bypass-for-password-accounts-cve-2026-123158 · Vulnerabilities · patched · since 2026-11-05 · CVE-2026-123158 When sshd is built without PAM (Alpine's default `openssh-server`, or `openssh-server-pam` with `UsePAM no`), it checks a password by running `crypt()` on the attempt and comparing the result with the stored field. An account whose field is too short to hold a salt, such as `*` (no password), gets the salt of another account with a `$5$`/`$6$` hash. sshd has done this since a 2016 timing-attack mitigation (`pick_salt()`). musl's SHA-crypt rejects keys over 256 bytes and returns the literal string `*` as its error value. So a password of roughly 257–1024 bytes "hashes" to `*`, matches the stored field, and logs the attacker in on the first try. Affected hosts need: `PasswordAuthentication yes`, any account with a `*` password field, and at least one local account with a real `$5$`/`$6$` hash. Accounts with `*` include `root` (shipped as `root:*` since about Alpine 3.16, but blocked by the default `PermitRootLogin prohibit-password`), key-only users, directory users from nslcd, winbind or sssd, and the `alpine` user with passwordless `doas` in tiny-cloud images 3.18.11–3.24.3. Every Alpine release from 3.16 to 3.24 and edge is affected. OpenSSH's maintainers say the root cause is outside sshd and shipped a workaround in 10.6p2 (2026-11-05); Alpine backported a patch to every branch back to 3.16 and cut point releases 3.21.10, 3.22.8, 3.23.8 and 3.24.4. The reporter is Özgün Kültekin of Wakefield, who calls the bug "sshole". Public lab/PoC repositories appeared within hours. There are no reports of in-the-wild exploitation so far. Hosts not affected: locked `!` accounts, `UsePAM yes`, and glibc-based distros. Fixed: OpenSSH portable 10.6p2; Alpine openssh: edge 10.6_p2-r0 · 3.24 10.3_p1-r2 · 3.23 10.2_p1-r1 · 3.22 10.0_p1-r11 · 3.21 9.9_p2-r1 · 3.20 9.7_p1-r6 · 3.19 9.6_p1-r3 · 3.18 9.3_p2-r4 · 3.17 9.1_p1-r7 · 3.16 9.0_p1-r6; Alpine point releases 3.21.10, 3.22.8, 3.23.8, 3.24.4 (and images built from them) Do: Upgrade and restart sshd: `apk upgrade -U -a && rc-service sshd restart`. Upgrading the package does NOT restart the running daemon.; Confirm the fixed openssh package for your branch: edge 10.6_p2-r0, 3.24 10.3_p1-r2, 3.23 10.2_p1-r1, 3.22 10.0_p1-r11, 3.21 9.9_p2-r1, 3.20 9.7_p1-r6, 3.19 9.6_p1-r3, 3.18 9.3_p2-r4, 3.17 9.1_p1-r7, 3.16 9.0_p1-r6. Non-Alpine musl/no-PAM builds: upgrade to OpenSSH 10.6p2.; If you cannot patch now, set `PasswordAuthentication no`, run `sshd -t`, restart sshd, and check that `sshd -T | grep -i passwordauthentication` prints `no`.; Hunt for prior use: search `/var/log/messages` (or your syslog) for `Accepted password for` on accounts that have no password (root, alpine, key-only and directory users). Review doas/sudo use, new users, authorized_keys and recently changed files. Treat a host as compromised if you cannot rule a login out.; Rebuild Alpine-based VM and cloud images (tiny-cloud qcow2/VHD/AMI 3.18.11–3.24.3 ship an `alpine` user with `*` and passwordless doas) and any container images that run sshd with password authentication.; Inventory other musl/no-PAM sshd builds (OpenWrt/Buildroot/Yocto images that ship OpenSSH rather than Dropbear, postmarketOS) and lock no-password accounts with `!` instead of `*` where possible. ### [critical] Apache Shiro: DefaultLdapRealm accepts any password with auth caching enabled (CVE-2026-129349) https://helix.ml/threat-watch/i/apache-shiro-defaultldaprealm-accepts-any-password-with-auth-caching-enabled-cve-2026-129349 · Vulnerabilities · patched · since 2026-11-14 · CVE-2026-129347, CVE-2026-129348, CVE-2026-129349 Apache disclosed three Shiro flaws on 2026-11-14. The critical one, CVE-2026-129349 (CVSS 4.0 9.1 as published by Apache on oss-sec), affects DefaultLdapRealm and JndiLdapRealm only when authenticationCachingEnabled is set to true, which is not the default. After an account has logged in successfully once, the cached authentication path skips the LDAP bind, and the AllowAllCredentialsMatcher that the realm installed by default accepts any password, so anyone who knows that username can sign in as it. Apache fixed it in 3.0.1 by no longer installing that matcher by default; 3.0.1 was tagged on GitHub on 2026-08-23, about twelve weeks before the advisory, and its release notes do not name the CVE. The two companion low-severity issues (CVE-2026-129347 for empty stored credentials matching empty passwords, CVE-2026-129348 for PropertiesRealm reloads that keep removed users and narrowed permissions) are fixed in 3.0.2, tagged 2026-11-08. No exploitation or public proof of concept was found, and NVD had not yet analysed the CVEs at research time. Action: upgrade to 3.0.2; if on the LDAP realm, keep authentication caching off or configure a credentials matcher that checks the cached data. Fixed: 3.0.1 (CVE-2026-129349); 3.0.2 (CVE-2026-129347, CVE-2026-129348) Do: Upgrade Apache Shiro to 3.0.2 or later (3.0.1+ covers the LDAP realm flaw); If you use DefaultLdapRealm/JndiLdapRealm with authenticationCachingEnabled=true, disable caching or configure SimpleCredentialsMatcher immediately; Avoid storing empty credentials and reject empty passwords at the application layer; If using PropertiesRealm, verify reloads correctly remove deleted users and narrowed permissions ### [high] Telegram widgets: suspected in-the-wild exploitation drops WebRAT (Salat Stealer) via PowerShell https://helix.ml/threat-watch/i/telegram-widgets-rce-webrat-salat-stealer-telegram-suspected-in-the-wild-exploitation-of-new-widgets-feature-to-run-powe · Vulnerabilities · partial · since 2026-11-13 Kaspersky warned on 2026-11-13 that a flaw in Telegram's widgets feature, which Pavel Durov promoted on 2026-11-10, is presumably being exploited: users who visited certain Telegram channels had a PowerShell script run without their knowledge, which fetched and started WebRAT (Salat Stealer), a Rust trojan giving the operator control of the machine that can also steal cryptocurrency-wallet and Telegram account data. Kaspersky published no technical details and no count of targeted users. Separately, researchers at V12 posted a video of a message-triggered RCE on Android (a Calculator window opened) and claimed a second exploit against the patched Windows client; V12 says running the QuickJS engine unsandboxed inside messages creates a large attack surface. Telegram said the Android issue was fixed on 2026-11-10 and the fix was deployed to all users on 2026-11-11. Durov disputed the reports, saying an attack required the user to start a third-party service and that this path is fully patched server-side. Kaspersky advises not launching widgets until a full fix ships, and none of the sources documents a fixed Windows build. Do: Do not launch or interact with Telegram widgets until an official fix is released; Update Telegram clients (Android and Windows) to the latest available versions; Hunt for unexpected PowerShell execution and WebRAT/Salat Stealer indicators on endpoints used with Telegram; Block or monitor cryptocurrency wallet and Telegram session data exfiltration paths; Restrict Telegram usage on sensitive systems until the issue is fully resolved ### [high] Dropbear: per-destination forwarding limits bypassed via unix sockets, docker.sock reachable https://helix.ml/threat-watch/i/dropbear-permitopen-permitlisten-bypass-via-unix-socket-forwarding-docker-sock-reachable · Vulnerabilities · unpatched · since 2026-11-13 A researcher's oss-sec post dated 2026-11-13 says Dropbear's unix-socket forwarding (direct-streamlocal and streamlocal-forward) checks only the coarse no-port-forwarding flag and never the permitopen or permitlisten lists that TCP forwarding enforces, so an authorized_keys entry restricted to one TCP destination can still open unix sockets. The author tested only HEAD commit 99baf182 (2026-11-11), built from source on a local rig. The test key was limited to permitopen 127.0.0.1:1; after the account was added to the docker group, the same key reached /var/run/docker.sock and got a Docker API response, while a non-permitted TCP destination was refused. Reach to Docker therefore depends on host configuration (the account must be able to use the socket), not on the key alone. The author describes commit 796a89ef (2024-04-03) as an incomplete fix. The commit page shows it changes src/svr-tcpfwd.c and the manpage, not src/svr-streamfwd.c as the write-up says. No CVE, vendor advisory, upstream response or fixed version was found, and the Dropbear 2026.93 release notes do not mention this behavior. The author's suggested mitigation is to refuse streamlocal channels for keys that carry per-destination restrictions. The test harness and logs are not published in the post. Do: Audit authorized_keys for keys carrying permitopen/permitlisten restrictions on Dropbear servers and treat them as unrestricted for unix-socket forwarding until patched; Remove unix-socket forwarding capability for restricted accounts (e.g. disable DROPBEAR_SVR_LOCALSTREAMFWD/REMOTESTREAMFWD at build time or restrict via server options); Restrict OS-level access to sensitive unix sockets (docker.sock, database sockets) for accounts holding restricted SSH keys — e.g. remove from docker group; Monitor for direct-streamlocal@openssh.com and streamlocal-forward@openssh.com channel requests in Dropbear logs; Track upstream Dropbear for a fix adding path-based permit restrictions in svr-streamfwd.c ### [critical] Apache HTTP Server mod_http2 RCE and mod_vhost_alias/mod_md flaws disclosed (CVE-2026-119618 et al.) https://helix.ml/threat-watch/i/apache-http-server-mod-http2-rce-and-mod-vhost-alias-mod-md-flaws-disclosed-cve-2026-119618-et-al · Vulnerabilities · patched · since 2026-11-11 · CVE-2026-106023, CVE-2026-109163, CVE-2026-119618, CVE-2026-63292 On 2026-11-11 the Apache HTTP Server project released httpd 2.4.71, which fixes three CVEs disclosed the same day on oss-security. The most serious is CVE-2026-119618, rated critical by Apache and CVSS 9.8 by Red Hat. It is a thread-safety flaw in mod_http2 assign_headers(). For a body-less HTTP/2 request that has Content-Type but no Content-Length, a worker thread writes a synthesised "Content-Length: 0" into the stream's shared header table instead of the request's private copy. When that table is full, the write takes memory from the connection's lock-free allocator while the connection thread is using it, which corrupts the child process heap. Red Hat says public research has shown a reliable exploit against a default-configured server with mod_http2 enabled. The flaw affects httpd 2.4.55 through 2.4.70 and standalone mod_h2 2.0.8 through 2.0.42, and is reachable without authentication. Only servers that negotiate HTTP/2 through the Protocols directive (h2 or h2c) are affected; the httpd default is http/1.1. Red Hat rates RHEL 9 and 10 Important, because their mod_http2 builds give each stream its own allocator, and says RHEL 8 and earlier are not affected. A public reproduction lab (googlesky/CVE-2026-119618) appeared on 2026-11-12. It demonstrates heap corruption under ASan and TSan, not a working RCE exploit. The fix is a one-line change in httpd 2.4.71. Standalone mod_h2 has no fixed release, because the repository was archived in June 2026 at v2.0.42. Debian has the fix only in sid (2.4.71-1); bookworm, trixie and forky are still marked vulnerable. The same release fixes CVE-2026-106023 (important, mod_md NULL pointer dereference crash, 2.4.30 through 2.4.70) and CVE-2026-109163 (low, mod_vhost_alias one-byte stack overwrite, 2.4.69 and 2.4.70, a follow-up to CVE-2026-63292). No in-the-wild exploitation has been reported. Fixed: Apache HTTP Server 2.4.71 (released 2026-11-11; fixes CVE-2026-119618, CVE-2026-106023 and CVE-2026-109163); Debian sid apache2 2.4.71-1 Do: Upgrade Apache HTTP Server to 2.4.71 (released 2026-11-11), which fixes CVE-2026-119618, CVE-2026-106023 and CVE-2026-109163. Before relying on a distro package, confirm it carries the fix: on 2026-11-12 Debian stable (bookworm, trixie) was still unpatched.; If you cannot patch now and do not need HTTP/2, remove h2 and h2c from every Protocols directive, including those inside VirtualHost blocks, and reload. The PoC author verified this workaround in their lab. Servers that serve only http/1.1 (the httpd default) are not affected.; Find HTTP/2-enabled httpd first: run `apachectl -M | grep http2`, grep the configuration for `Protocols.*h2`, and check internet-facing hosts with `curl -sI --http2 https://host` (it reports HTTP/2 when h2 is negotiated).; Standalone mod_h2 users have no fixed module release (repository archived, last tag v2.0.42). Move to httpd 2.4.71's bundled mod_http2. Disregard third-party claims of a mod_h2 2.0.43.; RHEL 8 and earlier are not affected. RHEL 9 and 10 are rated Important; apply the httpd/mod_http2 errata as Red Hat ships them.; Patch internet-facing servers that negotiate HTTP/2 first. A public lab now shows exactly which requests trigger the corruption. ### [high] PostgreSQL 14-18 update: 52 CVEs fixed, many allow code execution by a database user https://helix.ml/threat-watch/i/postgresql-bulk-security-update-fixes-52-cves-incl-many-rce-flaws · Vulnerabilities · patched · since 2026-11-12 · CVE-2026-116303, CVE-2026-116304, CVE-2026-116306, CVE-2026-116307, CVE-2026-116308, CVE-2026-116309, CVE-2026-116310, CVE-2026-116311, CVE-2026-116312, CVE-2026-116313, CVE-2026-116314, CVE-2026-116315, CVE-2026-116316, CVE-2026-116317, CVE-2026-116318, CVE-2026-116319, CVE-2026-116320, CVE-2026-116321, CVE-2026-116322, CVE-2026-116323, CVE-2026-116325, CVE-2026-116326, CVE-2026-116327, CVE-2026-116328, CVE-2026-116329, CVE-2026-116330, CVE-2026-116331, CVE-2026-116332, CVE-2026-116333, CVE-2026-116335, CVE-2026-116336, CVE-2026-120139, CVE-2026-120141, CVE-2026-120142, CVE-2026-120143, CVE-2026-120144, CVE-2026-120145, CVE-2026-120146, CVE-2026-120147, CVE-2026-120148, CVE-2026-120149, CVE-2026-120150, CVE-2026-120151, CVE-2026-120152, CVE-2026-120153, CVE-2026-120154, CVE-2026-14661, CVE-2026-16237, CVE-2026-16240, CVE-2026-18447, CVE-2026-6460, CVE-2026-6467 On 12 November 2026 the PostgreSQL Global Development Group released 18.7, 17.12, 16.16, 15.20 and 14.25, plus 19 RC 2, fixing 52 security vulnerabilities and over 140 bugs. About 30 of the CVEs are memory-safety or type-confusion flaws scored CVSS 8.8 that can run code as the operating-system user running the database; most need a database account that can create objects or run SQL, and no CVE scores 9.0 or above. Other fixes cover SQL injection in pg_dump, reindexdb and extension scripts, privilege bypasses in row security, DROP OWNED and REASSIGN OWNED, and a dblink/subscription file-append issue. The only unauthenticated item is a denial-of-service restart (CVE-2026-116322, CVSS 5.3). PostgreSQL 14 is end-of-life with this release. No exploitation is reported. Some fixes need post-update steps (reindexing, new backups for incremental backups), so read the release notes before upgrading. Fixed: 18.7; 17.12; 16.16; 15.20; 14.25; 19 RC 2 Do: Upgrade all PostgreSQL deployments to 18.7, 17.12, 16.16, 15.20 or 14.25 (or 19 RC 2 for testing); prioritise databases where untrusted users can create objects or run SQL; Plan migration off PostgreSQL 14, which is now end-of-life; Review the release notes and apply required post-update steps (reindexing, new full backup before incremental backups); Restrict CREATE privileges and untrusted SQL execution, and avoid restoring untrusted dump or archive files, until patched; Rebuild or restore workflows using patched pg_dump, pg_restore and pg_basebackup clients ### [critical] Citrix NetScaler November 2026 patches: Console auth bypass and ADC/Gateway memory flaws (CVE-2026-120869 et al.) https://helix.ml/threat-watch/i/citrix-netscaler-november-2026-patch-batch-12-flaws-incl-auth-bypass-and-dos-cve-2026-120868-et-al · Vulnerabilities · patched · since 2026-11-12 · CVE-2026-107407, CVE-2026-107408, CVE-2026-107409, CVE-2026-107410, CVE-2026-107412, CVE-2026-107415, CVE-2026-118268, CVE-2026-118269, CVE-2026-120864, CVE-2026-120868, CVE-2026-120869 Citrix shipped fixes on 2026-11-12 for a batch of NetScaler and Secure Access Client flaws. Press coverage counts 12 flaws; the full per-CVE list is only partly confirmed because the Citrix bulletin was not readable as text. The one flaw in NVD so far is CVE-2026-120869, an authentication bypass in NetScaler Console: NVD says an unauthenticated attacker can bypass authentication controls and reach restricted resources (CVSS v4.0 9.3 in NVD, 9.4 in press reports). The NetScaler ADC and Gateway memory-corruption flaws are rated 9.3 to 9.5 in press lists and are described as causing denial of service or unpredictable behavior. No source reviewed describes any of them as remote code execution; press summaries say CVE-2026-107412 and CVE-2026-120868 need no login, and neither is in NVD yet. Two local privilege-escalation flaws in Secure Access Client, CVE-2026-118268 (Windows, to SYSTEM) and CVE-2026-118269 (Linux, to root), are rated 8.5 and need a low-privileged local account. Citrix reports no known exploitation and no public proof of concept has been confirmed. None of the listed IDs is in the CISA KEV catalogue snapshot of 2026-11-10, which predates the batch, so that check is point-in-time only. The Hacker News reports that Citrix attributes the fixes to AI-assisted code review; that detail could not be re-checked against the bulletin. Patch NetScaler ADC and Gateway to 14.1-73.57 or 13.1-64.39 (FIPS builds 14.1-73.58 FIPS, FIPS/NDcPP builds 13.1-37.293), NetScaler Console to 14.1-74.34, and Secure Access Client to 26.11.1.6 (Windows; one advisory mirror shows 26.11.16, so confirm against the bulletin) or 26.10.4 (Linux). Fixed: NetScaler ADC/Gateway 14.1-73.57+, 13.1-64.39+; NetScaler ADC FIPS 14.1-73.58 FIPS+; NetScaler ADC FIPS/NDcPP 13.1-37.293 FIPS & NDcPP+; NetScaler Console 14.1-74.34+; Secure Access Client Windows 26.11.1.6+; Secure Access Client Linux 26.10.4+ Do: Upgrade NetScaler ADC and NetScaler Gateway to 14.1-73.57 or later, or 13.1-64.39 or later. FIPS builds need 14.1-73.58 FIPS or later; FIPS/NDcPP builds need 13.1-37.293 or later.; Upgrade NetScaler Console to 14.1-74.34 or later, and restrict network access to it until the upgrade is done, because CVE-2026-120869 is an unauthenticated authentication bypass.; Upgrade Secure Access Client for Windows to 26.11.1.6 or later and for Linux to 26.10.4 or later.; Review DNS/DoH, NetScaler Push, Email OTP and ICA proxy configurations on ADC and Gateway, and limit their exposure until patched.; After patching, check internet-facing appliances for signs of compromise from before the upgrade. Citrix reports no known exploitation, so treat this as a precaution, not a finding. ### [critical] N-able N-central: unauthenticated root RCE and Windows SYSTEM exploits work on Hotfix 2 (CVE-2026-90436) https://helix.ml/threat-watch/i/4everable2shell-agent2shell-n-able-n-central-new-root-rce-and-windows-system-exploits-work-against-patched-servers-cve-2 · Vulnerabilities · partial · since 2026-11-12 · CVE-2026-90436 N-able is investigating a second set of N-central flaws, published by researcher Sean Bales days after it shipped 2026.4 Hotfix 2 (build 2026.4.1.26) for the earlier Able2Shell set. He says two new proofs of concept, 4EverAble2Shell and Agent2Shell, still work against that build. 4EverAble2Shell gives unauthenticated remote code execution as root on the N-central server, but only where the Legacy API is turned on and reachable from the internet, which N-able says is not the default. Agent2Shell chains more than ten zero-days to run commands as NT AUTHORITY\SYSTEM on managed Windows agents if the attacker knows the target appliance ID. Previdian reports probing from 157.20.241[.]16 within hours of the PoC release. The observed chain used the Apache Axis 1.x flaw CVE-2026-90436 for initial access, then a root step, then a curl download of a shell script. No successful exploitation is confirmed. Hosted environments are not affected by the server issue. The reporting reviewed names no vendor hotfix for the new flaws. N-able urges customers to disable the Legacy API. Fixed: 2026.4.1.26 (Hotfix 2) fixes seven of the 33 Able2Shell flaws (see the Able2Shell record); no fix announced for 4EverAble2Shell or Agent2Shell in the reporting reviewed; CVE-2026-90436: upstream commit 41f3b4e8 to apache/axis-axis1-java; Axis 1.x is end-of-life and the project will not release a 1.x fix (per NVD) Do: Disable the Legacy API on every internet-exposed N-central on-premise server now (Administration > Mail and Network Settings > Network Security > LEGACY APIS). N-able says it is off by default and not recommended.; Keep N-central at 2026.4.1.26 or later and apply further N-able updates as they ship. Hotfix 2 does not stop the reported PoCs, per the researcher.; CVE-2026-90436 (Apache Axis 1.x): the project will not release a 1.x fix. Apply the upstream commit 41f3b4e8 or migrate to Axis 2, and confirm whether the N-central SOAP stack depends on Axis 1.x.; Block and monitor traffic to and from 157.20.241[.]16 and watch for probe activity against N-central; Restrict N-central and Legacy API exposure to the public internet via firewall or VPN; Review managed Windows agents for signs of SYSTEM-level command execution ### [high] Windows CLFS: unpatched SignDrop zero-day grants SYSTEM and reportedly bypasses tamper protection https://helix.ml/threat-watch/i/signdrop-windows-clfs-authentication-mitigation-zero-day-grants-system-privileges-and-edr-tamper-protection-bypass · Vulnerabilities · unpatched · since 2026-11-10 On 2026-11-10 a researcher publishing as Nightmare Eclipse (MSNightmare) posted a public proof-of-concept called SignDrop, with a write-up, for a local privilege escalation in the Windows Common Log File System (CLFS) driver; BleepingComputer covered it on 2026-11-12. As the researcher describes it, the bug abuses the CLFS logfile-authentication mitigation that Microsoft began shipping in updates from 2025-10-28: it makes a privileged process such as Windows Defender open a log, which causes the .cnpf authentication file to be created with SYSTEM rights while inheriting the original file's permissions, and he says this also gets past antimalware tamper protection. The PoC needs code execution as a standard user first, so it is a second-stage tool rather than a remote entry point. The researcher reports it works on Windows 11 26H2 and Windows Server 2025 builds, and his claim that about 80% of Windows installations are exposed is an unmeasured estimate. Third-party reports are mixed. Mallory relays that Howler Cell (Cyderes) reproduced a SYSTEM shell on 2026-11-12 and that 78researchlab found it works but is noisy, since every UAC prompt opened both a standard-user and a SYSTEM shell; we could not retrieve either firm's own post. The tamper-protection bypass is a claim that no source we read has verified; Shield53 repeats it without testing. As of the latest sources (2026-11-13) there is no CVE, no CVSS score, no Microsoft statement or patch, and no CISA KEV entry, and no in-the-wild exploitation has been reported. Microsoft's documentation says the mitigation's Mode registry value 0 is enforcement and 1 is learning, and that disabling it (value 2) is not recommended. Do: Check HKLM\SYSTEM\CurrentControlSet\Services\CLFS\Authentication\Mode. Microsoft Learn documents 0 = enforcement, 1 = learning and 2 = disabled by an administrator; the researcher and ThreatDefence treat 0 and 1 as the exposed states (not independently verified); Do not run the public PoC on production or corporate endpoints; treat it as hostile code and block or alert on its execution; Apply Microsoft's patch as soon as it is released; monitor for the fix; Keep the CLFS authentication mitigation enabled: Microsoft Learn says disabling it is not recommended, and the researcher says disabling it would re-expose systems to CLFS memory-corruption attacks; Monitor privileged processes (e.g., Defender) for unexpected file creation in log directories and watch for tamper-protection bypass attempts; Restrict which processes can open CLFS log files where feasible ### [critical] Red Hat Quay: unauthenticated service-key takeover forges registry tokens (CVE-2026-123710) https://helix.ml/threat-watch/i/red-hat-quay-key-server-flaw-lets-attackers-forge-registry-tokens-and-steal-private-container-images-cve-2026-123710 · Vulnerabilities · partial · since 2026-11-11 · CVE-2026-110697, CVE-2026-123710 Red Hat rates CVE-2026-123710 Critical (CVSS 3.1 9.8) in the service key server of Red Hat Quay. Key rotation accepts any JWT signed by an already-approved service key that carries the expected audience and issuer, and it never checks that the token is a rotation credential. An anonymous token from /v2/auth passes that check, so a caller who can reach the /keys endpoint can install a signing key of its own, with no expiry and inherited approval. Registry tokens signed with that key are then accepted for every repository, which allows reading and overwriting private images and tags, and the rogue key survives restarts. Red Hat's exposure condition is that /keys is reachable by untrusted clients, and anonymous access is one route to a token rather than the only one. Red Hat lists Quay 3 and Quay 5 as affected. SecurityWeek reports no exploitation in the wild and no public proof-of-concept. The fix is merged to the upstream quay/quay master branch (PR #8056, 2026-11-11), but no tagged release contains it as of 2026-11-12, and Red Hat names no fixed build for this CVE. The same merge fixes CVE-2026-110697, a separate authenticated flaw (CVSS 9.9) that lets a namespace-name wildcard expose other tenants' robot tokens; Red Hat names fixed builds 3.9.28 through 3.18.3 for that one only. Until a fixed build ships, restrict access to /keys, audit service_key_rotate and service_key_delete events, and remove any unrecognised approved key. Do: Restrict network access to the /keys service-key endpoint at the proxy or load balancer so only trusted internal components can reach it; Review audit events service_key_rotate and service_key_delete, especially any with a null expiration_date, and remove every approved service key you do not recognise; Expect reinstalled keys: deleting a rogue key is not durable until a fixed build is deployed, because the anonymous request can install a new one. Watch for disrupted pushes and pulls; Treat images and secrets in affected registries as potentially exposed, and rotate credentials embedded in images; Apply the fixed Quay build as soon as one ships. The fix for CVE-2026-123710 is on master (quay/quay PR #8056) but is not in any tagged release as of 2026-11-12 ### [critical] PTC Windchill/FlexPLM: critical deserialization RCE patched (CVE-2026-126325) https://helix.ml/threat-watch/i/ptc-windchill-flexplm-critical-deserialization-rce-patched-cve-2026-126325 · Vulnerabilities · patched · since 2026-11-12 · CVE-2026-12569, CVE-2026-126325 PTC published a patch on 2026-11-11 for CVE-2026-126325, a critical remote code execution flaw in Windchill and FlexPLM caused by deserialization of untrusted data (CVSS v4 9.2, CVSS v3.1 9.0). PTC applied the fix itself to PTC-hosted instances; self-hosted customers have to apply it, following eSupport article CS483249. That article needs a PTC login and could not be read for this review, so the affected and fixed release lines are not confirmed here. Neither the PTC notice nor the press coverage reviewed says the flaw is exploited or that a public proof of concept exists, and it is not in CISA KEV. The same product line carries CVE-2026-12569, which is in KEV with known ransomware use and was exploited by Cl0p-linked actors to drop JSP web shells and take product data. PTC's trust center also lists a separate RCE, CVE-2026-77645, and an SSRF, CVE-2026-77646, from August 2026. The trust center does not say whether this patch relates to either. Do: Apply the PTC patch for CVE-2026-126325 now if you self-host Windchill or FlexPLM (PTC eSupport CS483249; a Windchill administrator applies it). Confirm the fixed build for your release line in that article.; Monitor hosts for code execution, backdoor access and JSP web shells. PTC's trust center lists hunt patterns for CVE-2026-12569 webshells under /Windchill/login/.; Apply network restrictions to limit exposure if patching is not immediately possible; Review whether the environment was exposed to CVE-2026-12569 (KEV, ransomware use) and hunt for compromise indicators ### [critical] Zoom Workplace for Windows: unauthenticated privilege escalation (CVE-2026-108186, CVE-2026-108187) https://helix.ml/threat-watch/i/asus-router-firmware-flaws-zoom-workplace-for-windows-privilege-escalation-cve-2026-108186 · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-108186, CVE-2026-108187 Zoom published two bulletins on 2026-11-10 for Zoom Workplace for Windows. CVE-2026-108186 (CVSS 9.6 per Zoom and NVD) is an improper input validation flaw that lets an unauthenticated attacker over the network escalate privilege. CVE-2026-108187 (CVSS 8.1) is an origin validation error with the same effect. Both CVSS 3.1 vectors require user interaction, so the attack needs a victim action. CVE-2026-108186 affects versions from 7.1.0 up to 7.1.8; CVE-2026-108187 affects versions before 7.1.5. Version 7.1.8 and later carries fixes for both. Neither Zoom nor SecurityWeek reports exploitation, neither CVE is in the CISA KEV catalog (checked against the 2026.11.10 feed), and NVD still lists both as awaiting analysis. Fixed: Zoom Workplace for Windows 7.1.8 or later (fixes both CVEs) Do: Update Zoom Workplace for Windows to 7.1.8 or later, which carries the fixes for both CVEs; Inventory Windows endpoints running Zoom Workplace below 7.1.8, since the affected range starts at 7.1.0 ### [high] Windows 10 KB5131096 ESU: November 2026 security fixes, plus PowerShell and NTLM signing changes https://helix.ml/threat-watch/i/windows-10-kb5131096-esu-november-2026-patch-tuesday-fixes-incl-actively-exploited-outlook-zero-day · Vulnerabilities · patched · since 2026-11-10 Microsoft's KB5131096, released 2026-11-10, is the cumulative update for Windows 10 version 22H2 under Extended Security Updates and for Windows 10 Enterprise and IoT Enterprise LTSC 2021. It moves builds to 19045.7846 and 19044.7846. The KB's change list covers an Active Directory printer-install failure (error 87), Intune control of the trusted .rdp publisher thumbprint policy, RDS shadow-connection failures on Windows Server 2016 and 2012 R2, a Cuba time-zone rule change, and the remaining USB Audio Class 1.0 symptoms from September. Two changes can break working setups. PowerShell signature validation can now reject scripts that were accepted before because they are not valid UTF-8 with a byte order mark, so they must be re-saved and re-signed. NTLM message signing is now offered by default for non-anonymous connections, which can affect legacy clients. The package also carries Microsoft's November security fixes for these Windows 10 builds; by our reading of Microsoft's CVRF, 30 of the CVEs mapped to it are rated Critical and none is marked exploited. The actively exploited Outlook zero-day, CVE-2026-124194, belongs to Microsoft 365 Apps and Office LTSC 2021 and 2024 and is not in this package. Fixed: 19045.7846; 19044.7846 Do: Install KB5131096 (builds 19045.7846 and 19044.7846) on Windows 10 22H2 ESU and Enterprise or IoT Enterprise LTSC 2021 devices. It requires servicing stack 19044.6088 or later, and the combined SSU KB5132689 ships with it; Patch Outlook separately: CVE-2026-124194 is fixed in the November Microsoft 365 Apps and Office LTSC 2021/2024 updates, not in KB5131096. Until then, warn users about unexpected message files and consider blocking outbound NTLM to untrusted destinations; Re-save PowerShell scripts as UTF-8 with a byte order mark and sign them again before relying on signature validation after this update; Audit for non-anonymous NTLM connections that are not signed, since signing is now offered by default, and test legacy clients before broad rollout; After installing, verify AD-published printer installs, Intune RDP publisher policy, RDS shadow connections where used, and USB Audio Class 1.0 devices; For Dynamic Update installs, include boot.stl on the installation media, or secure boot validation can fail with 0xc0430001 ### [critical] N-able N-central: 33 flaws with public PoCs, unauthenticated XXE-to-RCE; 7 of 33 fixed in Hotfix 2 https://helix.ml/threat-watch/i/able2shell-n-able-n-central-30-zero-days-published-hosted-environments-taken-offline · Vulnerabilities · partial · since 2026-11-09 · CVE-2026-125878, CVE-2026-125879, CVE-2026-125880, CVE-2026-125881, CVE-2026-125882, CVE-2026-125883, CVE-2026-91000 A researcher publishing as Oracle-Security put proof-of-concept exploits for 33 alleged zero-day flaws in N-able's N-central RMM on GitHub on 8-9 November 2026, all tested on pre-release appliance 2026.4.0.31. Press reports name the researcher as Sean Bales; N-able does not name him. The worst chain needs no credentials: an unauthenticated XXE in service import can reset an account, run code as the nable user, and add an enabled Default Administrator. N-able took Hosted N-central offline on 9 November as a precaution and told on-premises customers to keep their servers offline until upgraded. On 10 November it released N-central 2026.4 Hotfix 2 (build 2026.4.1.26), which fixes seven CVEs: CVE-2026-125883, CVE-2026-125878, CVE-2026-125879, CVE-2026-125880, CVE-2026-125881, CVE-2026-125882 and CVE-2026-91000 (CVSSv4 8.3 to 9.3). N-able has not said which of the other 26 flaws the hotfix covers, and a follow-up hotfix has been promised with no date. N-able says hosted servers are upgraded and back in service, and that it has no evidence of customer exploitation, while its status page tracks scanning tied to the published code. Beazley and watchTowr report exploitation attempts against N-central honeypots, and Previdian reports attempts against a decoy; none of these is a confirmed customer compromise. Five older N-central CVEs in CISA KEV (including CVE-2026-86218) are separate; Tenable says it is not known whether Able2Shell relates to them. Apply Hotfix 2 to on-premises servers (agents and probes first), check authentication activity from 8 November onward, and restrict access to service-import endpoints until the follow-up fix ships. Fixed: 2026.4.1.26 (N-central 2026.4 Hotfix 2, released 2026-11-10; fixes seven of the 33 flaws) Do: Upgrade on-premises N-central to 2026.4 Hotfix 2 (build 2026.4.1.26), agents and probes first; N-able says keep an on-premises server offline until it is upgraded; Treat Hotfix 2 as partial: N-able has not mapped the other 26 flaws to a fix and the follow-up hotfix has no date; track N-able's status page and apply it when it ships; Until then, restrict N-central exposure and block unauthenticated access to service-import (SOAP) endpoints; Check login and authentication activity from 2026-11-08 onward, and look for new admin accounts, unexpected script execution and Take Control sessions on managed endpoints; Review firewall and edge logs for XXE and SOAP-endpoint probes; the source IPs and endpoints are published by Beazley Security Labs and Previdian; Review downstream customer environments that N-central manages for unauthorized scripts or access; earlier N-central exploitation was seen spraying scripts into managed environments ### [critical] MongoDB Server: unauthenticated use-after-free in replica set catch-up, crash or code exec (CVE-2026-118383) https://helix.ml/threat-watch/i/cve-2026-118383-mongodb-server-unauthenticated-use-after-free-in-replica-set-primary-catch-up-can-crash-or-execute-code · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-118383 MongoDB published CVE-2026-118383 on 2026-11-10, together with fixed releases on each branch from 4.4 to 9.0 that the vendor lists as affected. The bug is a use-after-free in the replication code that runs when a newly elected replica set primary finishes catching up on writes it missed. The vendor's CVE text says requests that need no authentication can reach it, and that the outcome is a server crash or, under certain conditions, arbitrary code execution. Datadog Security Labs reproduced the crash on 8.0.34 in a four-member lab: two unauthenticated ping requests carrying readConcern.afterOpTime left stale waiters on one member, and a failover that made that lagging member primary segfaulted it. The same two requests were rejected with Unauthorized on 8.0.37. The attacker cannot create the lag or the election, so the crash needs a member to fall behind and then win an election, which is why the CVSS attack complexity is High. Datadog did not demonstrate code execution. Standalone mongod is not affected. MongoDB said it was not aware of exploitation when it announced the fix, and the CISA KEV catalogue has no entry for this CVE. Fix: upgrade every replica set member, including shard and config server members and members offline during maintenance, to the fixed release for its branch, then confirm the running versions. Until then, limit network access to MongoDB to the systems that need it; that reduces who can send the requests but does not fix the bug. Atlas clusters are upgraded automatically during maintenance, according to MongoDB. Fixed: 9.0.7; 8.3.16; 8.0.37; 7.0.48; 6.0.30; 5.0.35; 4.4.32 Do: Upgrade every replica set member (including shards, config servers, containers, and members offline during maintenance) to the fixed release for your branch and verify running versions; Until upgraded, restrict network access to MongoDB to only the systems that need it; Monitor MongoDB logs for COMMAND 'Slow query' events (ID 51803) with command.ping, readConcern.afterOpTime, and MaxTimeMSExpired/MaxTimeMSExpired errName indicating planted waiters; If deferring maintenance (e.g. on Atlas), apply the upgrade promptly ### [medium] 'Smart Eye' Overseas Chinese Affairs platform: leaked files describe UFWD AI profiling of diaspora talent https://helix.ml/threat-watch/i/overseas-chinese-affairs-big-data-platform-smart-eye-ccp-ai-system-scrapes-diaspora-data-to-scout-spies-in-taiwan-chip-i · Vulnerabilities · no-fix-needed · since 2026-11-11 NetAskari, a research group, says internal documents leaked to it describe the Overseas Chinese Affairs Big Data Platform, which it calls 'Smart Eye', run from Jinan University's Big Data Center and attributed to the United Front Work Department. Its published analysis says the platform uses big-data analysis of social media, news and other open sources, including AI-supported web scraping, to build profiles of overseas Chinese and to pick 'fruitful' targets among foreign academics and think-tank staff. Dark Reading reports that the platform scraped Instagram, Facebook, X, Weibo and WeChat through rotating foreign proxies and produced profiles of ethnic-Chinese chip-sector talent, including Skymizer executives, Taiwanese university students and a Microsoft AI research center director, with approaches through hometown, alumni and professional networks. Those specifics are not in the NetAskari excerpt and were not checked against Dark Reading. The leaked documents are not published, and the NetAskari excerpt does not say how they were obtained or verified. This is influence and recruitment activity rather than a technical breach, so no patch applies. Do: Review insider-risk and foreign-influence exposure for staff with ties to Chinese diaspora networks, especially in semiconductor and AI roles; Harden social media privacy practices for executives and researchers in strategic industries; Brief leadership and security teams on UFWD recruitment and influence tactics (hometown/alumni/professional association approaches); Monitor for anomalous data-scraping infrastructure and proxy networks targeting your organization's online presence; Coordinate with government counterintelligence channels if approached by suspected influence actors ### [high] pappl-retrofit: path traversal in PPD add/delete lets attackers write or delete files (CVE-2026-123747) https://helix.ml/threat-watch/i/cve-2026-123747-pappl-retrofit-directory-traversal-in-add-delete-ppdfiles-actions-lets-attackers-write-delete-arbitrary · Vulnerabilities · partial · since 2026-11-11 · CVE-2026-123747 pappl-retrofit, OpenPrinting's library for turning classic CUPS PPD drivers into Printer Applications, does not sanitise filenames in its add-ppdfiles and delete-ppdfiles actions. Each name is joined to the PPD storage directory as-is, so ../ sequences let a client write or delete files outside it. The GitHub advisory (GHSA-wxj6-x3wg-4vmc, published 2026-11-11, CVE-2026-123747) includes a working curl proof of concept, and says an uploaded PPD with a cupsFilter setting can also lead to command execution. Affected: versions up to and including 1.0b2. Authentication is contested: NVD and the advisory text describe the service as unauthenticated and running as root by default, but the advisory's CVSS vector requires low privileges. Red Hat's rating (CVSS 7.3, Important) assumes high privileges and user interaction, and says default RHEL 10, CentOS Stream 10 and Fedora installs are not exposed because the storage directory /var/lib/legacy-printer-app/ppd is not created by default. The advisory scores the flaw 9.6 (critical). NVD and Red Hat, the only independent scorers, rate it 7.3 (high). No in-the-wild exploitation is reported, and the flaw is not in CISA's KEV catalogue (the copy checked predates disclosure). The fix is commit c65a28f on master. No release containing it is listed in the sources checked, so upgrade once one ships. Do: Check whether pappl-retrofit (<= 1.0b2) is installed, including as the legacy-printer-app service, and whether /var/lib/legacy-printer-app/ppd exists; Restrict network access to pappl-retrofit's web and IPP interfaces until a fixed release is installed; If building from source, apply fix commit c65a28f on master; Upgrade to the first release that includes c65a28f (none is listed in the sources checked); Check logs for add-ppdfiles or delete-ppdfiles requests with ../ in the filename ### [critical] Xen Orchestra: crafted cookie logs in as a token owner with admin rights (VSA-2026-062) https://helix.ml/threat-watch/i/xen-orchestra-auth-bypass-vsa-2026-062-vates-attackers-bypass-auth-via-crafted-cookie-take-over-xcp-ng-hosts-as-root-cve · Vulnerabilities · unknown · since 2026-11-10 Vates' security advisory VSA-2026-062 (rated Critical, no CVE assigned) describes an authentication bypass in Xen Orchestra and the Xen Orchestra Appliance. The REST cookie parser turns a crafted authenticationToken cookie (an empty JSON object) into an empty object, and the token lookup treats that as an unfiltered query and returns the first unexpired login token, so the request is authenticated as that token's owner with admin rights. Anyone who can reach the management API can then list reusable login token IDs and act as an administrator. The vendor states the flaw is being actively exploited by an unnamed actor, and warns that this can lead to code execution on managed pools and root takeover of XCP-ng hosts. The exploitation leaves no log entries, so detection depends on HTTP inspection and the published indicators. The vendor names Xen Orchestra 6.9.1 (stable) and 6.10.0 (latest) as fixed. The upstream fix (PR #10748) was merged on 2026-11-10, after the 6.10.0 release date, so the fixed release has not yet been confirmed from a public changelog. Discovered by TDC NET Security (Filip Magic, Christoffer Olsen). Fixed: Xen Orchestra 6.9.1 (stable), 6.10.0 (latest); xo-server 5.211.2, 5.213.1 Do: Upgrade Xen Orchestra and XOA to a build that contains the VSA-2026-062 fix (advisory: 6.9.1 stable, 6.10.0 latest; xo-server 5.211.2 / 5.213.1), then confirm the running xo-server version. Do not treat 6.10.0 as fixed until a release containing PR #10748 is confirmed; If patching is delayed, keep Xen Orchestra on a dedicated management network, not the internet, and block 91.211.251.112 (port 8001) and the listed onion endpoint; Add WAF or reverse-proxy rules for authenticationToken=j:{} and authenticationToken=j:%7B%7D; the advisory warns that regex rules can be bypassed with other encodings; Hunt for the IOCs: authenticationToken=j:{} in HTTP logs; a VM named pivot2; hvm_serial: file VM configs; /etc/cron.d/zz-ustar and zz-ustar2, /.ustar5, /.ustar6 and /etc/ld.so.preload on XCP-ng hosts. Absence of IOCs does not prove a host is clean; If compromise is suspected, rotate all passwords and secrets used by the platform, including proxy, backup, S3, root and user accounts ### [critical] Plone: anonymous visitor can run code on the server via CMFEditions modifier and sandbox escapes https://helix.ml/threat-watch/i/plone-rce-chain-via-cmfeditions-modifier-and-allow-module-sandbox-escapes-anonymous-attackers-can-run-code-on-server · Vulnerabilities · patched · since 2026-11-10 The Plone/Zope Security Team published five GitHub security advisories and fixed releases on 2026-11-10 for a chained remote code execution flaw. It affects Plone 6.0, 6.1 and 6.2 and older releases back to 4.3. An anonymous visitor can start the chain: missing security declarations on CMFEditions portal_modifier classes let them change a versioning TALES condition, which runs when an editor saves versioned content, so the code executes with that editor's privileges. Overly broad allow_module declarations in CMFPlone, MimetypesRegistry and plone.batching expose imported modules to restricted code, and plone.restapi (rated Medium) lets restricted code bypass permissions through json. The chain was reported by Mokrane ABDELMALEK via Ghent University's VDP. The advisory says the details are now fully public and that a working exploit may not take long to build. It reports no in-the-wild exploitation. Fixed releases: CMFPlone 6.2.3, 6.1.6, 6.0.16; CMFEditions 5.0.2, 4.0.7; MimetypesRegistry 4.0.2, 3.0.2; plone.batching 3.0.3, 2.0.8; plone.restapi 10.2.1, 9.15.8. For Plone 4.3 through 6.2 the alternative is Products.PloneHotfix20261110. Until patched, block URLs containing portal_modifier at the web server and keep the block after upgrading. Restart all Zope instances after updating. Check for changed modifier conditions with @@check-modifier-conditions as Manager, or the hotfix's startup log line 'POSSIBLE ATTACK: MODIFIED VERSIONING CONDITIONS FOUND'. A changed condition persists in the database after the fix, and the advisory says to treat any condition you do not recognise as a compromise. Fixed: Products.CMFPlone 6.2.3/6.1.6/6.0.16; Products.CMFEditions 5.0.2/4.0.7; Products.MimetypesRegistry 4.0.2/3.0.2; plone.batching 3.0.3/2.0.8; plone.restapi 10.2.1/9.15.8; Products.PloneHotfix20261110 (Plone 4.3–6.2) Do: Update to the fixed package versions for your Plone series, or install Products.PloneHotfix20261110 and restart all Zope instances; Block URLs containing portal_modifier at the web server as a workaround and keep it after updating; Check for attacks: use @@check-modifier-conditions as Manager, or review hotfix startup logs for 'POSSIBLE ATTACK: MODIFIED VERSIONING CONDITIONS FOUND'; treat unrecognized conditions as compromise; Review custom Python Scripts/templates for imports now blocked, declaring needed names via ModuleSecurityInfo rather than allow_module; Upgrade out-of-support Plone 5.2 and older to Plone 6 ### [high] Microsoft Outlook: NTLM hash-leak zero-day exploited in Nov 2026 Patch Tuesday (CVE-2026-124194) https://helix.ml/threat-watch/i/microsoft-november-2026-patch-tuesday-fixes-record-1-235-flaws-actively-exploited-outlook-ntlm-hash-leak-zero-day-cve-20 · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-112982, CVE-2026-114155, CVE-2026-114171, CVE-2026-116717, CVE-2026-116718, CVE-2026-116719, CVE-2026-116720, CVE-2026-116721, CVE-2026-116722, CVE-2026-116922, CVE-2026-116926, CVE-2026-116927, CVE-2026-116929, CVE-2026-116930, CVE-2026-116931, CVE-2026-116932, CVE-2026-116933, CVE-2026-116934, CVE-2026-116935, CVE-2026-116936, CVE-2026-116937, CVE-2026-116938, CVE-2026-116939, CVE-2026-116940, CVE-2026-116941, CVE-2026-116942, CVE-2026-116943, CVE-2026-116944, CVE-2026-116945, CVE-2026-116946, CVE-2026-118117, CVE-2026-118118, CVE-2026-118119, CVE-2026-118120, CVE-2026-118148, CVE-2026-118288, CVE-2026-118292, CVE-2026-118293, CVE-2026-118294, CVE-2026-118295, CVE-2026-118296, CVE-2026-118297, CVE-2026-118298, CVE-2026-118299, CVE-2026-118300, CVE-2026-118301, CVE-2026-118302, CVE-2026-118700, CVE-2026-118701, CVE-2026-118702, CVE-2026-118703, CVE-2026-118704, CVE-2026-118705, CVE-2026-118706, CVE-2026-118707, CVE-2026-118708, CVE-2026-118709, CVE-2026-118710, CVE-2026-118711, CVE-2026-118712, CVE-2026-118713, CVE-2026-118714, CVE-2026-118715, CVE-2026-118716, CVE-2026-118717, CVE-2026-118718, CVE-2026-118719, CVE-2026-118720, CVE-2026-118740, CVE-2026-118741, CVE-2026-118742, CVE-2026-118746, CVE-2026-118747, CVE-2026-119382, CVE-2026-119824, CVE-2026-119886, CVE-2026-119981, CVE-2026-119990, CVE-2026-120796, CVE-2026-121681, CVE-2026-122162, CVE-2026-122163, CVE-2026-122164, CVE-2026-122165, CVE-2026-122166, CVE-2026-122167, CVE-2026-122168, CVE-2026-122169, CVE-2026-122170, CVE-2026-122171, CVE-2026-122172, CVE-2026-122173, CVE-2026-122174, CVE-2026-122175, CVE-2026-122297, CVE-2026-122474, CVE-2026-122475, CVE-2026-122987, CVE-2026-124194, CVE-2026-85929 Microsoft's November 2026 Patch Tuesday fixes 1,094 vulnerabilities dated to the release day, per Microsoft's own CVRF document and SecurityWeek; BleepingComputer's 1,235 figure (which it says excludes 108 earlier fixes) does not reproduce from the CVRF. The only CVE marked Exploitation Detected is CVE-2026-124194, a spoofing flaw in Outlook: a crafted message file makes Outlook connect to an attacker-controlled network location without the usual external-content warning, which can expose the user's NTLM authentication hash so an attacker can impersonate them. The attack needs network reach and a user to open the file. CISA added CVE-2026-124194 to the Known Exploited Vulnerabilities catalog on 2026-11-10 with a federal due date of 2026-11-24. Microsoft rates it Moderate (CVSS 3.1 6.5, user interaction required). Deploy the November updates, prioritising Outlook, warn users about unexpected message files, and consider blocking outbound NTLM authentication to untrusted destinations. Do: Deploy the November 2026 Patch Tuesday updates immediately, prioritizing the Outlook CVE-2026-124194 fix on all clients; Apply Windows 11 KB5132061 and KB5131087 cumulative updates; Warn users not to open unexpected message/attachment files and report phishing; Consider blocking outbound SMB/NTLM authentication to untrusted destinations and enforcing NTLM blocking policies; Review the MSRC update guide for other affected components (Azure File Sync, Telephony Service, Remote Desktop Client) and patch accordingly ### [high] Google Chrome: XML use-after-free zero-day exploited in the wild (CVE-2026-125350) https://helix.ml/threat-watch/i/chrome-xml-use-after-free-zero-day-exploited-in-the-wild-cve-2026-125350 · Vulnerabilities · patched · since 2026-11-09 · CVE-2026-125342, CVE-2026-125343, CVE-2026-125344, CVE-2026-125345, CVE-2026-125346, CVE-2026-125347, CVE-2026-125348, CVE-2026-125349, CVE-2026-125350, CVE-2026-126840, CVE-2026-126841, CVE-2026-126843, CVE-2026-126844, CVE-2026-126845, CVE-2026-126847, CVE-2026-126848, CVE-2026-126849 Google released Chrome Stable 157.0.8103.50/.51 for Windows and Mac and 157.0.8103.50 for Linux on 2026-11-09, with nine security fixes. Google says it is aware that an exploit for CVE-2026-125350, a high-severity use-after-free in the XML component, exists in the wild; it is the only fix in the release that Google marks as exploited, and Google has not described the attacks. The bug was reported to Google by an anonymous researcher on 2026-11-06. The other eight fixes are six High and two Medium issues in WebGL, FileSystem, Compositing, GPU, V8 and XML. CVE-2026-125346, a second XML use-after-free reported on 2026-10-12, is one of them, and Google does not say it is exploited. Press coverage calls this the eighth exploited Chrome zero-day of 2026; the seven earlier ones have their own records. Extended Stable 156.0.8078.96 for Windows and Mac shipped the same day with the same fix, and Chrome for Android 157.0.8103.50 carries the desktop fixes. CISA had not added the CVE to KEV as of catalog 2026.11.09. Fixed: 157.0.8103.50/.51 (Windows, Mac); 157.0.8103.50 (Linux, Android); 156.0.8078.96 (Extended Stable, Windows, Mac) Do: Update Chrome to 157.0.8103.50/.51 (Windows, Mac) or 157.0.8103.50 (Linux) and restart the browser; Extended Stable users update to 156.0.8078.96; Android users update via Google Play to 157.0.8103.50; Force browser updates across managed fleets via enterprise update policies and verify versions; Review CISA KEV additions for CVE-2026-125350 and monitor for related exploitation indicators; Ensure Chromium-based browsers (Edge, Brave, etc.) are updated once they merge the underlying fixes ### [high] Apache TomEE: 38 CVEs fixed in 10.3.0/11.0.0-M2 incl. EJB auth bypass on JDK 21+ (CVE-2026-86336) https://helix.ml/threat-watch/i/apache-tomee-38-cves-fixed-in-10-3-0-11-0-0-m2-incl-ejb-auth-bypass-on-jdk-21-cve-2026-86336 · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-111239, CVE-2026-111240, CVE-2026-111241, CVE-2026-111242, CVE-2026-111243, CVE-2026-111244, CVE-2026-111245, CVE-2026-111246, CVE-2026-111247, CVE-2026-111248, CVE-2026-111249, CVE-2026-111250, CVE-2026-111251, CVE-2026-111256, CVE-2026-111257, CVE-2026-111258, CVE-2026-111259, CVE-2026-111260, CVE-2026-111261, CVE-2026-111262, CVE-2026-111263, CVE-2026-111264, CVE-2026-86336, CVE-2026-86531, CVE-2026-86532, CVE-2026-86811, CVE-2026-87755, CVE-2026-87756, CVE-2026-87758, CVE-2026-97849, CVE-2026-97867, CVE-2026-97870, CVE-2026-97872, CVE-2026-97874, CVE-2026-97880, CVE-2026-97881, CVE-2026-97888, CVE-2026-97890 On 10 November 2026 Apache TomEE published 38 security advisories, consolidated on oss-security on 12 November. TomEE 10.3.0 and 11.0.0-M2 fix 26 of them: five rated Important by the project (EJB method authorization skipped on JDK 21+ (CVE-2026-86336); MicroProfile JWT role checks not enforced for JAX-RS resources (CVE-2026-111244); well-known administrative credentials created at startup in the default configuration with no users defined (CVE-2026-111257); empty-password LDAP bind accepted where the directory allows unauthenticated binds (CVE-2026-111261); and an EJB JAX-WS endpoint auth bypass via alternative HTTP methods (CVE-2026-111264)). The other 21 fixed items are rated Moderate or Low. Twelve more affect optional modules (embedded OpenEJB HTTP server, Hessian remoting, SSH admin shell) that are removed in both releases and have no fixed version. Several fixes change default behaviour, so read the migration guides. No exploitation is reported. Fixed: 10.3.0; 11.0.0-M2 Do: Upgrade to Apache TomEE 10.3.0 or 11.0.0-M2 and read the 10.3.0 and 11.0.0-M2 migration guides, since several fixes change default behaviour; If upgrading is not possible, stop using or block network access to the embedded OpenEJB HTTP server, Hessian remoting and the SSH administration shell from untrusted clients; these modules have no fix; On JDK 21+, review EJB method authorization in deployments (CVE-2026-86336) and confirm checks are enforced after upgrade; Check for default administrative users on startup, LDAP identity stores that accept unauthenticated binds, and EJB web service endpoints with authentication configured (CVE-2026-111257, CVE-2026-111261, CVE-2026-111264) ### [critical] Ivanti EPM SQLi RCE (CVE-2026-121653) and CSA auth bypass (CVE-2026-121683) in a 19-flaw Patch Tuesday https://helix.ml/threat-watch/i/citrix-ivanti-patch-tuesday-19-flaws-incl-critical-csa-auth-bypass-cve-2026-121683-and-epm-sqli-rce-cve-2026-121653 · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-102705, CVE-2026-105425, CVE-2026-121653, CVE-2026-121683 On Tuesday 2026-11-10 Ivanti and Citrix published patches for 19 vulnerabilities. Ivanti fixed 13 flaws across five products. The two criticals are CVE-2026-121653 (CVSS 9.8 per NVD), an unauthenticated SQL injection in the Endpoint Manager Core that leads to remote code execution, fixed in EPM 2024 SU8; and CVE-2026-121683 (CVSS 9.4 per NVD), a missing-authentication flaw in Cloud Services Application 5.0.8 that lets an unauthenticated remote attacker add and delete trusted EPM Core certificate entries, fixed in CSA 5.0.8.1. SecurityWeek reports that Dual NIC deployments cannot be exploited externally. Ivanti also fixed CVE-2026-105425 in Sentry (high, CVSS 7.2 per NVD), an OS command injection that NVD says needs administrator privileges; the same CVE is fixed in EPMM, which also received five medium issues. Citrix fixed six XenServer flaws, including a high-severity CVE-2026-102705 that SecurityWeek says allows Pool Secret inference and host compromise; that CVE does not appear in NVD or in Citrix's bulletin CTX697170, so it is unverified. XenServer also received updates for AMD CPU issues. Neither vendor reports exploitation. None of these CVE IDs is in CISA KEV at catalog 2026.11.10. Patch CSA to 5.0.8.1 and EPM to 2024 SU8 first, then Sentry and EPMM to the versions listed below. Fixed: CSA 5.0.8.1; EPM 2024 SU8; Sentry 11.0.0, 10.8.3 and 10.7.4 (CVE-2026-105425); EPMM 12.11.0.0, 12.10.0.1 and 12.9.0.3 (CVE-2026-105425) Do: Patch Ivanti CSA to 5.0.8.1 immediately; verify Dual NIC exposure as a temporary mitigation; Update Ivanti EPM to 2024 SU8 and restrict database access pending patch; Update Sentry to 11.0.0, 10.8.3 or 10.7.4, and EPMM to 12.11.0.0, 12.10.0.1 or 12.9.0.3 (CVE-2026-105425; NVD says admin privileges are needed); Apply Citrix XenServer security updates, including AMD CPU mitigations; Review certificate trust entries in EPM Core for tampering and audit for signs of exploitation ### [critical] Fortinet Patch Tuesday: 89 advisories, ~17 critical, incl. FortiClient Fortinet2Shell code execution via CORS https://helix.ml/threat-watch/i/fortinet2shell-fortinet-patch-tuesday-fortinet-89-advisories-fix-100-flaws-incl-critical-rce-and-auth-bypass-across-fort · Vulnerabilities · patched · since 2026-11-10 On Tuesday 2026-11-10 Fortinet published 89 PSIRT advisories fixing nearly 100 flaws, about 17 of them critical and roughly 80 high. FortiMail has the largest share of the criticals, followed by FortiAuthenticator, FortiOS, FortiManager and FortiAnalyzer, with critical fixes also in FortiSASE and FortiClient. The criticals include code execution, authentication bypass, account takeover, SSRF, IBE webmail session compromise and super-admin command execution. The headline FortiClient bug, which researcher Kevin Joensen of Baldur Security named Fortinet2Shell, is a CORS misconfiguration that lets an attacker's web page reach a local service and run code on a victim who visits it, but only when FortiPAM is enabled. No CVE or Fortinet advisory for that specific bug was retrieved; the detail rests on SecurityWeek's report of the researcher's statement. Fortinet's own statements in the batch say no exploitation: the FortiGuard pages for FG-IR-26-267 and FG-IR-26-207 both read Known Exploited: No, SecurityWeek says Fortinet makes no mention of exploitation, and no CVE from this batch is in CISA KEV at catalog 2026.11.10. Fixes are in the PSIRT advisories. Patch the FortiMail, FortiAuthenticator and FortiOS criticals first, patch FortiClient and check whether FortiPAM is enabled, and restrict management interfaces to trusted networks. Do: Immediately review Fortinet's PSIRT advisories and patch FortiMail, FortiAuthenticator, FortiOS, FortiManager and FortiAnalyzer critical issues first; Patch FortiClient and review whether FortiPAM is enabled; treat the Fortinet2Shell flaw as high priority; Restrict access to Fortinet management interfaces and local services from untrusted web content; Monitor Fortinet systems for signs of exploitation before and after patching; Prioritise upgrades quickly given AI is shortening disclosure-to-exploitation windows ### [critical] Fortinet patch batch: FortiOS OSPF auth bypass, FortiAuthenticator hard-coded password & auth flaws https://helix.ml/threat-watch/i/fortinet-patch-batch-fortios-ospf-auth-bypass-fortiauthenticator-hard-coded-password-auth-flaws · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-126214, CVE-2026-126215, CVE-2026-126216, CVE-2026-126217, CVE-2026-126244, CVE-2026-126251, CVE-2026-126253, CVE-2026-126257 Fortinet's PSIRT published five advisories on 2026-11-10 for fixes that the earlier Fortinet record (2026-10-26) tracks as shipped in its 2026-10-22 hardening releases. The highest-rated is FG-IR-26-267 (CVE-2026-126253, CVSS 9.3, adjacent-network attacker): an authentication bypass in the FortiOS and FortiSASE OSPF daemon that lets an unauthenticated attacker broadcast crafted LSAs. FG-IR-26-271 (CVE-2026-126257, CVSS 9.0, CWE-259) is a hard-coded password in FortiAuthenticator. FG-IR-26-265 (CVE-2026-126251, CVSS 9.1, CWE-863) is an incorrect-authorization flaw in FortiAuthenticator that the advisory says may allow authentication bypass. FG-IR-26-258 (CVE-2026-126244, CVSS 7.5, CWE-770) is an unbounded-resource denial-of-service in FortiAuthenticator; its title says fastapi, but the body and product table name only FortiAuthenticator. FG-IR-26-234 (CVE-2026-126214 to CVE-2026-126217, CVSS 8.7) covers FortiAnalyzer, FortiManager and FortiSASE-Sovereign. Each advisory names fixed builds per branch. Fortinet reports no exploitation, none of the eight CVEs is in CISA KEV, and no virtual patch is offered. Three advisories credit Claude Mythos with the discovery. Fixed: FortiOS 8.0.2, 7.6.8, 7.4.13, 7.2.14 or above (FG-IR-26-267); FortiSASE 26.3.2 for 26.3.1 and 26.3.1_a; older FortiSASE branches: migrate to a fixed release (no version named); FortiAuthenticator 8.0.4 or above, 6.6.11 or above (FG-IR-26-271, FG-IR-26-265, FG-IR-26-258); FortiAnalyzer and FortiManager 8.0.2, 7.6.8, 7.4.12, 7.2.13 or above (FG-IR-26-234); FortiSASE-Sovereign: migrate to a fixed release (no version named) Do: Upgrade FortiOS to 8.0.2, 7.6.8, 7.4.13 or 7.2.14 or later (FG-IR-26-267), prioritising internet-reachable and OSPF-enabled FortiGate devices; Upgrade FortiSASE 26.3.1 and 26.3.1_a to 26.3.2; migrate older FortiSASE branches to a fixed release; Upgrade FortiAuthenticator to 8.0.4 or 6.6.11 or later (FG-IR-26-271, FG-IR-26-265, FG-IR-26-258); Upgrade FortiAnalyzer and FortiManager to 8.0.2, 7.6.8, 7.4.12 or 7.2.13 or later (FG-IR-26-234); migrate FortiSASE-Sovereign to a fixed release; Restrict OSPF and FortiAuthenticator management interfaces to trusted networks until patched; Review FortiAuthenticator logs for signs of authentication anomalies ### [critical] FortiMail critical flaws: unauth command injection, SQL injection, IBE session hijack; FortiOS sshd auth bypass https://helix.ml/threat-watch/i/fortimail-critical-flaws-unauth-command-injection-sql-injection-ibe-session-hijack-fortios-sshd-auth-bypass · Vulnerabilities · unknown · since 2026-11-10 · CVE-2026-126148, CVE-2026-126155 On 2026-11-10 Fortinet published four PSIRT advisories. FortiMail is affected by three unauthenticated critical flaws: command injection via crafted HA join-request packets (CVE-2026-126148, CVSS 9.8, allows arbitrary code execution), SQL injection via crafted requests (CVSS 9.4), and incorrect authorization in IBE webmail (CVE-2026-126155, CVSS 9.1, lets an attacker obtain an authenticated session as any IBE recipient and read or send their encrypted mail). FortiMail 7.2.0 through 8.0.1 are affected. Separately, a path traversal in the ssh daemon of FortiOS, FortiSASE, FortiProxy and FortiPAM (CVSS 7.4) allows authentication bypass via crafted credentials. Two flaws were found internally with AI model assistance. No in-the-wild exploitation is reported; fixes are available — upgrade promptly. Do: Upgrade FortiMail to a fixed release immediately, prioritising internet-facing IBE webmail and HA clusters; Upgrade FortiOS, FortiProxy, FortiPAM and FortiSASE sshd-affected builds to patched versions; Restrict management and HA interfaces to trusted networks until patched; Review FortiMail IBE mailboxes and logs for signs of unauthorised session access or sent mail; Monitor Fortinet PSIRT for updated fixed-version details, as advisories list affected versions but not fixed builds in these feeds ### [critical] GitLab CE/EE: Git merge heap overflow may allow command execution in Gitaly (CVE-2026-105437) https://helix.ml/threat-watch/i/gitlab-critical-patch-release-19-5-4-19-4-6-19-3-8-heap-overflow-in-bundled-git-enables-rce-in-gitaly-cve-2026-105437 · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-105437 GitLab's 2026-11-10 critical patch releases (19.5.4, 19.4.6, 19.3.8) update the Git binary bundled with GitLab CE/EE to fix CVE-2026-105437, a heap overflow in Git's xdiff three-way merge code. The merged-output size is summed in a 32-bit int, and the conflict-marker-size attribute, which can be set from a committed .gitattributes file, accepts any positive integer and multiplies the marker writes. A small file with a huge marker size and many conflicts can wrap that total, so Git writes past a buffer that is too small. Upstream Git describes the result as a crash with possible heap corruption. GitLab says the out-of-bounds writes could execute arbitrary commands inside the Gitaly environment; upstream says a forge is exposed when it reads attributes from repository trees (attr.tree or --attr-source) and any user who can push and open a merge request can trigger the overflow. On clients, any merge, rebase, cherry-pick or merge-tree in an untrusted repository that carries the attribute can trigger it. Self-managed CE/EE from 16.2 before 19.3.8, 19.4 before 19.4.6 and 19.5 before 19.5.4 are affected; GitLab.com and GitLab Dedicated are already patched. Official packages and images ship the patched Git; self-compiled installs must rebuild Gitaly's Git and the Git at git.bin_path, or use Git 2.56.1 or a distribution patch. Upstream fixed it in Git 2.50.2, 2.51.3, 2.52.1, 2.53.1, 2.54.1, 2.55.1 and 2.56.1, and the vulnerable code is present in every release checked from v1.7.0 through v2.56.0. No exploitation and no public proof of concept were found, and CISA KEV has no entry. Reported via HackerOne by taise. Fixed: 19.5.4; 19.4.6; 19.3.8; Upstream Git 2.50.2; Upstream Git 2.51.3; Upstream Git 2.52.1; Upstream Git 2.53.1; Upstream Git 2.54.1; Upstream Git 2.55.1; Upstream Git 2.56.1 Do: Upgrade self-managed GitLab CE/EE to 19.5.4, 19.4.6 or 19.3.8 immediately; For self-compiled installations, rebuild Git binaries in Gitaly and the Git referenced by git.bin_path, or upgrade Git to 2.56.1 or a distribution-patched version; Verify no other bundled/embedded Git copies remain unpatched; Review Gitaly logs for signs of anomalous command execution; Until patched, servers and forges can add conflict-marker-size=7 to each repository's info/attributes (it overrides any in-tree .gitattributes), or turn off server-side mergeability checks for untrusted repositories (upstream workaround); Before merging, rebasing or cherry-picking in an untrusted repository, check its .gitattributes for a very large conflict-marker-size, or add the same override to .git/info/attributes ### [critical] Adobe November 2026 Patch Tuesday: 110 vulnerabilities fixed across 9 products, incl. CVSS 10 Connect flaws https://helix.ml/threat-watch/i/adobe-november-2026-patch-tuesday-110-vulnerabilities-fixed-across-9-products-incl-cvss-10-connect-flaws · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-121349, CVE-2026-121350, CVE-2026-121351, CVE-2026-121352, CVE-2026-121353, CVE-2026-121354, CVE-2026-121355, CVE-2026-121356, CVE-2026-121887, CVE-2026-121889, CVE-2026-121890, CVE-2026-121892, CVE-2026-121894, CVE-2026-121896, CVE-2026-121898, CVE-2026-121900, CVE-2026-121902, CVE-2026-121904, CVE-2026-121906, CVE-2026-121908, CVE-2026-121910, CVE-2026-121912, CVE-2026-121913, CVE-2026-122102 Adobe's November 2026 Patch Tuesday fixes roughly 110 vulnerabilities across nine products, per SecurityWeek. The Adobe Connect bulletin APSB26-191 (published 2026-11-10, priority 2) covers 24 CVEs in Adobe Connect 12.12 and earlier on Windows. It rates five of them CVSS 10.0 (privilege escalation via SSRF, path traversal, incorrect authorization and CRLF injection). Twelve more are Critical between 9.1 and 9.9, covering arbitrary code execution, SQL injection, XXE, path traversal and stored XSS. The remaining seven are High or Medium, including an arbitrary file system read. SecurityWeek reports six CVSS 10 Connect flaws; the bulletin we could read shows five. Adobe states it is not aware of any exploits in the wild for any issue in this update. The other products in the release (AEM, Premiere Pro, Photoshop, After Effects, Audition, Dimension, Media Encoder and the Content Credentials SDK) are covered by bulletins not verified here. Update Adobe Connect first, then the desktop and SDK products as change control allows. Fixed: Adobe Connect 12.12.a; Adobe Connect 12.11.i; Adobe Connect 12.10.f; Adobe Connect 12.9.a; Adobe Connect 12.8.a (Windows; the bulletin does not map fixed versions to individual CVEs) Do: Prioritise updating Adobe Connect and AEM (priority 2, critical CVSS 10 flaws) as soon as change control allows; Update Premiere, Photoshop, After Effects, Audition, Dimension, and Media Encoder to the November 2026 releases; Update the Content Credentials Rust SDK / C2PA Tool if used in builds; Review internet-exposed Connect and AEM instances for signs of SSRF, path traversal, or injection attempts; Track Adobe advisories for any later exploitation reports and adjust patch priority accordingly ### [critical] SAP NetWeaver / Web Dispatcher: four CVSS 10.0 pre-auth flaws patched, incl. SAML bypass (CVE-2026-120017) https://helix.ml/threat-watch/i/sap-netweaver-web-dispatcher-four-cvss-10-0-pre-auth-flaws-fixed-on-patch-day-incl-rce-and-saml-auth-bypass-cve-2026-769 · Vulnerabilities · patched · since 2026-11-10 · CVE-2026-120017, CVE-2026-76964, CVE-2026-89103, CVE-2026-89126 On SAP's November 2026 Security Patch Day (10 Nov), SAP published four security notes, each rated CVSS 10.0 by SAP as CNA, and each flaw is reachable without authentication or user interaction. CVE-2026-76964 (ICM, Note 3778530) and CVE-2026-89126 (Web Dispatcher, Note 3778556) are memory-corruption bugs triggered by insufficiently validated network input, which SAP says can allow arbitrary code execution. CVE-2026-89103 (SAP Gateway, Note 3798345) is a default access control list that classifies an external attacker as an internal application server. CVE-2026-120017 (SAML2 service provider in AS ABAP, Note 3849308) does not sufficiently verify incoming authentication assertions under certain configurations, which lets an attacker impersonate arbitrary users, including privileged accounts. SAP's page lists the affected releases per note, including the HANA-related Web Dispatcher components HDB 2.00, XS_ADVANCED_RUNTIME 1.00 and SAP_EXTENDED_APP_SERVICES 1. Neither SAP's page nor NVD reports exploitation or a public proof of concept, and none of the four CVEs is in the CISA KEV feed checked on 2026-11-09. Patch each note for the components you run, and review SAML2 logons and Gateway access control, which are the paths these flaws describe. Fixed: SAP Notes 3778530 (ICM), 3778556 (Web Dispatcher), 3798345 (Gateway), 3849308 (SAML2 service provider); fixed release levels per note not verified Do: Patch SAP Note 3849308 (CVE-2026-120017) on AS ABAP systems that use SAML2 browser logon, and review trusted identity providers and SAML2 logon audit logs for privileged users logging on from unexpected sources; Patch SAP Note 3778530 (ICM, CVE-2026-76964) and SAP Note 3778556 (Web Dispatcher, CVE-2026-89126) together, and find every Web Dispatcher instance, including the HANA-related components listed on SAP's page; Patch SAP Note 3798345 (Gateway, CVE-2026-89103) and review Gateway access control lists, since the flaw is in default ACL handling; Check each system's release level against the affected lists on SAP's page, and confirm the fixed release level in each note, which this record did not verify ### [critical] CISA KEV additions: GPUStack, FreeSWITCH, OpenSSH, IBM Aspera Faspex flaws exploited in the wild https://helix.ml/threat-watch/i/cisa-kev-additions-gpustack-freeswitch-openssh-ibm-aspera-faspex-flaws-exploited-in-the-wild · Vulnerabilities · patched · since 2026-11-09 · CVE-2026-102895, CVE-2026-116779, CVE-2026-123158, CVE-2026-123218 CISA added four flaws to its Known Exploited Vulnerabilities catalog with a dateAdded of 2026-11-09 and a federal due date of 2026-11-12. CVE-2026-102895 (GPUStack, CVSS 9.8): before 2.2.0, a request that appeared to come from loopback was treated as the built-in administrator when force_auth_localhost was at its default of false, so a remote client could reach admin functions unauthenticated, including creating persistent admin API keys, reading worker and cloud-provider secrets, and running code through the inference-backend YAML endpoint; CISA and the GitHub advisory say exploitation was seen against a 2.1.1 deployment. CVE-2026-116779 (FreeSWITCH, CVSS 9.8): before 1.11.3, an inbound SIP MESSAGE could select an internal chat protocol and run FreeSWITCH API and OS commands; it needs an authenticated profile account, and the project's advisory says this path has been exploited in the wild. CVE-2026-123158 (OpenSSH, CVSS 8.1): before 10.6p2 on Linux built with musl libc and without PAM, sshd accepts a long password (about 256 to 1024 characters) for some account configurations; Alpine describes the condition as an account with a '*' password field plus another account with a $5$ or $6$ hash. CISA lists it as exploited, but I found no product-specific exploitation report. CVE-2026-123218 (IBM Aspera Faspex 5.0.0 to 5.0.19, CVSS 9.8): an authorized user can craft a message that bypasses authorization and gains full administrator rights; IBM fixed it in 5.0.19 PL1. CISA's KEV criteria require reliable evidence of active exploitation, so its listing counts as exploitation for this flaw too, although IBM's bulletin and the press coverage give no exploitation detail. Fixes exist for all four. Fixed: GPUStack 2.2.0; FreeSWITCH 1.11.3; OpenSSH 10.6p2; IBM Aspera Faspex 5.0.19 PL1 Do: Patch GPUStack to 2.2.0 or later; check for unauthorized admin API keys, changed users or models, edited configuration, and any reads of worker or cloud-provider secrets; Upgrade FreeSWITCH to 1.11.3 or later; if a SIP profile does not need chat, set enable-chat=false (the advisory says chat is on by default in the stock internal and external profiles); rotate credentials that the FreeSWITCH configuration could expose; Upgrade OpenSSH to 10.6p2 or later on Linux systems that use musl without PAM (Alpine is the documented example); until then set PasswordAuthentication no and restart sshd, because upgrading the package does not restart a running daemon; Upgrade IBM Aspera Faspex to 5.0.19 PL1 and confirm the faspex-core image reports 5.0.19 with fix pack 1, as IBM's bulletin specifies; Treat 2026-11-12 as the KEV due date for federal agencies; the KEV entries cite BOD 26-04, which carries forward the KEV criteria from BOD 22-01 ### [critical] Linux KVM: guest-to-host escape via malformed IPv4 checksum metadata lets guest VMs compromise the host (CVE-2026-99999) https://helix.ml/threat-watch/i/linux-kvm-guest-to-host-escape-via-malformed-ipv4-checksum-metadata-lets-guest-vms-compromise-the-host-cve-2026-99999 · Vulnerabilities · partial · since 2026-10-14 · CVE-2026-99999 PwnEscape (CVE-2026-99999) is a Linux host kernel flaw, despite the name: the bug is in the IPv4 stack, not KVM. A guest can hand the host virtio-net checksum metadata that points inside the IPv4 header; the host writes the completed checksum there, corrupting a header it has already validated and acted on. A guest that controls its own packets can therefore leak host kernel memory back to itself repeatedly, crash the host and every VM on it, and take over the host. On 5 November PWNAI published Part 2 with full proof-of-concept code: from inside a Firecracker v1.16.1 microVM on an unpatched Ubuntu 26.04 AWS host kernel (7.0.0-1011-aws), run without the jailer, it used the leak to find and hijack a vCPU thread and ran code as root on the host. The vulnerable code dates to 2.6.27 (2008). Paulos Yibelo of PWNAI found it in Vercel's sandbox challenge and reported it privately on 4 September; Michael Tsirkin's fix landed upstream on 14 October and is in stable 5.10.272, 5.15.223, 6.1.190, 6.6.159, 6.12.113, 6.18.57 and 7.2.11. Amazon Linux, Debian, SUSE, AlmaLinux and Red Hat 9/10 ship fixed kernels, but Ubuntu 22.04, 24.04 and 26.04 (including the linux-aws kernels) were still unfixed on 7 November, and there is no 7.0.y/7.1.y stable backport. AWS says EC2, Lambda and Fargate are not affected; self-managed Firecracker and KVM hosts are. No in-the-wild exploitation has been reported and the CVE is not in CISA KEV. Fixed: Upstream 6785011f8b16 ('ipv4: validate checksum_start before completing checksum'), mainline 7.3; Stable 5.10.272, 5.15.223, 6.1.190, 6.6.159, 6.12.113, 6.18.57, 7.2.11; Amazon Linux 2023 kernel-6.1.188-233.385.amzn2023, kernel6.12-6.12.110-135.201.amzn2023, kernel6.18-6.18.51-120.162.amzn2023; Amazon Linux 2 kernel-4.14.355-287.750, 5.4.302-229.494, 5.10.269-270.1108, 5.15.220-156.256; Debian bookworm 6.1.190-1 (DSA-6568-1), bookworm linux-6.12 6.12.113-1~deb12u1, trixie 6.12.115-2, sid 7.2.15-1; SUSE SLES 15 SP7 6.4.0-150700.53.87.3 (SUSE-SU-2026:4935-1), SLES 16.0 6.12.0-160000.40.1, SLE 12 SP5-LTSS 4.12.14-122.336.1; Ubuntu 26.10 linux 7.3.0-10.10 / linux-aws 7.3.0-1004.4; AlmaLinux 8 kernel-4.18.0-553.177.1.el8_10, 9 kernel-5.14.0-687.59.1.el9_8, 10 kernel-6.12.0-211.67.1.el10_2 Do: Patch every host that runs guests or sandboxes you do not fully trust (self-managed KVM, QEMU, Firecracker, Cloud Hypervisor): a public PoC now turns a guest into host root; Confirm the running kernel carries upstream 6785011f8b16 or its stable backport: 5.10.272, 5.15.223, 6.1.190, 6.6.159, 6.12.113, 6.18.57, 7.2.11 or 7.3+; check uname -r after the reboot; Distro builds: Amazon Linux 2023 kernel-6.1.188-233.385 / kernel6.12-6.12.110-135.201 / kernel6.18-6.18.51-120.162; Debian bookworm 6.1.190-1, trixie 6.12.115-2; SLES 15 SP7 6.4.0-150700.53.87.3; AlmaLinux 8/9/10 kernel-4.18.0-553.177.1.el8_10 / 5.14.0-687.59.1.el9_8 / 6.12.0-211.67.1.el10_2; Ubuntu 22.04/24.04/26.04 (generic and linux-aws, including 7.0.0-*-aws) has no fixed kernel yet: move untrusted-guest hosts to a fixed kernel (Ubuntu 26.10, or a stable kernel listed above) or a livepatch, or stop running untrusted guests there; Run Firecracker under the jailer with its default seccomp filters: Part 2 was demonstrated without the jailer, and the jailer narrows what a hijacked VMM thread can do (it does not fix the kernel bug); QEMU/libvirt hosts that cannot reboot: CloudLinux and Red Hat describe turning off virtio-net checksum offload (QEMU csum=off,guest_csum=off; libvirt host/guest csum='off'); untested against the bug, needs a guest restart and costs throughput. AWS states there is no workaround; On virtualization hosts, migrate or shut down guests before rebooting so the patch does not become an outage; Patch ordinary servers too: Red Hat says a local attacker can trigger it; until then, dropping CAP_NET_RAW from containers and setting user.max_user_namespaces=0 removes some local paths (Akamai, CloudLinux) ### [medium] Keycloak: manage-users admin escalates to manage-realm via composite role mapping (CVE-2026-96448) https://helix.ml/threat-watch/i/keycloak-manage-users-admin-escalates-to-manage-realm-via-composite · Vulnerabilities · unpatched · since 2026-09-25 · CVE-2026-96448 On 3 Nov 2026 Abraxas Labs published a lab proof of concept showing that a holder of manage-users, a delegated help-desk role that should not reach admin powers, can map a realm or client role onto their own account. That role's own name is not an admin role, but its composites include manage-realm or realm-admin. After the mapping, the same access token reads the full realm document and carries the nested admin role. A realm administrator must have built such a composite first. The defect is in the role-mapping permission check (RolePermissionsV2.canMapRole): for any non-admin role it returns true for manage-users and never walks the role's composites. We re-read that check at tags 26.8.1 and 26.7.5 and found it unchanged in both. Red Hat and NVD describe the same check as CVE-2026-96448, published 25 Sep 2026 and rated Moderate (CVSS 3.1 6.6); the write-up's 'no CVE' statement does not hold, and the write-up does not cite the CVE. Red Hat marks Red Hat Build of Keycloak as affected. We found no fixed upstream release and no community fix as of the 27 Sep 2026 analysis by Skycloak; no exploitation has been reported, and NVD's CISA SSVC enrichment lists exploitation as none and not automatable. The related CVE-2026-94000 (group-join composite walk, Red Hat Moderate) was closed upstream on 22 Sep 2026 as issue #53060; it is a sibling, not this defect. Until a fix ships, remove admin-including composites from roles that are not meant to grant admin, and cut manage-users down to the staff who need it. Do: Audit every realm role and client role for composites that include realm-management admin roles (manage-realm, realm-admin, manage-clients, manage-authorization and so on). Remove them from any role that is not itself meant to grant admin.; Until a fix ships, treat every manage-users holder as able to reach any admin power that sits inside such a composite. Cut manage-users down to the people who need it.; Hunt in admin events: look for role-mapping CREATE operations where the target user is the acting admin (resourcePath users//role-mappings/...), especially for composite roles.; Watch the Keycloak security advisories for a CVE and fixed release, and upgrade when one ships. ### [high] Keycloak: account takeover through SAML broker redirect-parameter split (no CVE yet) https://helix.ml/threat-watch/i/keycloak-account-takeover-through-saml-broker-redirect-parameter · Vulnerabilities · unpatched · since 2026-11-03 On 3 Nov 2026 Abraxas Labs published a lab-tested proof of concept for an unauthenticated account takeover in Keycloak's SAML identity-broker HTTP-Redirect endpoint. The researcher reproduced it on 26.8.1. An attacker who holds one genuine signed HTTP-Redirect response from a SAML identity provider that the realm trusts adds a forged SAMLResponse under a percent-encoded parameter name (SAML%52esponse). The endpoint parameter binding decodes that name and takes the forged document, while the signature check rebuilds the query from raw names and still accepts the genuine signature. Keycloak then signs in as the NameID inside the forged assertion, or creates that user through first-broker-login. The preconditions are a broker with signature validation on and 'Want Assertions signed' off, which is the usual result of metadata import, and an IdP that signs HTTP-Redirect responses; POST-only IdPs never give the attacker a signed GET. We re-read the Keycloak source at tags 26.8.1 and 26.7.5 and found the same code shape in both, so earlier 26.x lines are plausible but untested. As of 4 Nov 2026 we found no CVE for this flaw in NVD, no Keycloak advisory or GHSA, no fix release (the newest tag is 26.8.1), no vendor statement, and no report of exploitation in the wild; the only public PoC is the researcher's loopback lab. Until a fix ships, turn on Want Assertions Signed for every SAML broker, make IdPs POST-only where HTTP-Redirect is not needed, and look for logins whose NameID does not match the IdP session. Do: For every SAML identity-provider broker, turn on 'Want Assertions signed' (wantAssertionsSigned=true). Recheck brokers created by metadata import, which leaves it off. With it on, the forged unsigned document is rejected.; If an IdP does not need HTTP-Redirect responses, make it POST-binding only on the IdP side.; At the reverse proxy, reject GET requests to /realms/*/broker/*/endpoint whose query parameter names contain percent-encoding. Also reject any request carrying more than one SAMLResponse after decoding.; Hunt for abuse: look for IDENTITY_PROVIDER_LOGIN and first-broker-login user creations where the brokered NameID doesn't match the IdP's own sign-in logs for that session. Also search access logs for 'SAML%' in broker endpoint queries.; Watch the Keycloak security advisories for a CVE and fixed release, and upgrade when one ships. ### [critical] Linux kernel TCP use-after-free (CVE-2026-98432): Anthropic-built root exploit; Ubuntu and RHEL unfixed https://helix.ml/threat-watch/i/linux-kernel-tcp-use-after-free-race-mptcp-linked-root-exploit-ubuntu-rhel-unpatched-cve-2026-98432 · Vulnerabilities · partial · since 2026-10-11 · CVE-2026-121029, CVE-2026-121030, CVE-2026-121031, CVE-2026-121035, CVE-2026-121036, CVE-2026-121038, CVE-2026-121039, CVE-2026-121040, CVE-2026-121041, CVE-2026-121043, CVE-2026-121044, CVE-2026-121046, CVE-2026-121047, CVE-2026-121048, CVE-2026-121049, CVE-2026-121052, CVE-2026-121054, CVE-2026-121055, CVE-2026-121057, CVE-2026-121059, CVE-2026-121060, CVE-2026-121061, CVE-2026-121062, CVE-2026-121063, CVE-2026-121064, CVE-2026-121065, CVE-2026-121066, CVE-2026-121067, CVE-2026-121068, CVE-2026-121069, CVE-2026-121070, CVE-2026-121071, CVE-2026-121072, CVE-2026-121073, CVE-2026-121075, CVE-2026-121076, CVE-2026-121077, CVE-2026-121078, CVE-2026-121079, CVE-2026-121080, CVE-2026-121081, CVE-2026-121082, CVE-2026-121083, CVE-2026-121084, CVE-2026-121085, CVE-2026-121086, CVE-2026-121087, CVE-2026-121088, CVE-2026-121089, CVE-2026-121091, CVE-2026-74580, CVE-2026-98432, CVE-2026-99999 CVE-2026-98432 is a use-after-free race in the Linux kernel's IPv4 TCP connection setup, present since Linux 4.4. When a listener accepts a connection, the child socket takes over the request's IP options without copying them. Another CPU handling a retransmitted SYN can still be building a SYN-ACK from those options when the child socket frees them. Request-socket migration (net.ipv4.tcp_migrate_req, added in 5.14) has the same flaw. A remote, unauthenticated attacker who can reach any IPv4 TCP listener can try to win the race, and a lost race only fails the handshake, so the attacker can keep retrying. The fix, from Eric Dumazet and reported by Xinyang Ge of Anthropic, frees the options after an RCU grace period. It shipped in Linux 7.2.10 and 6.18.56 on 2026-10-11, in 6.12.113, 6.6.159, 6.1.190, 5.15.223 and 5.10.272 on 2026-10-15, and in 7.3. Anthropic says an exploit written by its model ran commands as root on a server with a multipath TCP (MPTCP) listener, with no credentials, taking about 53 minutes. It has published no exploit code. GitHub repositories created since 2026-10-29 claim race triggers, but none has been verified. Neither in-the-wild exploitation nor a CISA KEV listing has been reported. The kernel CNA scores it CVSS 9.8, and Red Hat rates it Critical: its RHSB-2026-016 bulletin warns of remote code execution, while its CVE page calls a crash or memory corruption the primary impact. Oracle rates it Important (8.1). Debian, Amazon Linux, SUSE 16 and Oracle UEK have shipped fixed kernels. As of 2026-10-30, Ubuntu (16.04 to 26.10) and RHEL 8, 9 and 10, including OpenShift and RHEL CoreOS, were still unfixed. Go 1.24 and later turn on MPTCP for Linux listeners by default. Fixed: Upstream: commit e3f33b0a1d89 (posted 2026-10-01, merged 2026-10-05, in 7.3-rc7); Linux 7.3 and later; Upstream stable: 7.2.10, 6.18.56 (2026-10-11); 6.12.113, 6.6.159, 6.1.190, 5.15.223, 5.10.272 (2026-10-15); Debian: bookworm linux 6.1.190-1 (DLA-4841-1) and linux-6.12 6.12.113-1~deb12u1 (DLA-4850-1); trixie 6.12.113-1 (DSA-6568-1); sid 7.2.10-1; Oracle Linux UEK: ELSA-2026-500397, ELSA-2026-500398, ELSA-2026-500399 (2026-10-16; OL7 to OL10); Amazon Linux: ALAS2023-2026-3335 (kernel6.12, 2026-10-29), ALAS2023-2026-3365 and ALAS2-2026-4065 (2026-10-30); SUSE 16 kernel packages at 6.12.0-160000.40.1 or later Do: Install a fixed kernel and reboot, Kubernetes nodes and container hosts included. Upstream fixes are in 7.3, 7.2.10, 6.18.56, 6.12.113, 6.6.159, 6.1.190, 5.15.223 and 5.10.272. Debian, Amazon Linux, SUSE 16 and Oracle UEK have shipped fixes. Ubuntu and RHEL 8, 9 and 10 have not as of 2026-10-30.; Until then, drop incoming IPv4 TCP packets that carry IP options, using Red Hat's nftables rule from RHSB-2026-016 ('ip protocol tcp ip hdrlength > 5 drop'). Red Hat says this reduces exposure but does not replace patching. The rule does not survive a reboot, and traffic between containers on the same host can bypass it.; Keep IPv4 TCP listeners off untrusted networks where you can, and patch internet-facing ones first, especially hosts with MPTCP listeners; Disable MPTCP where it is not needed (sysctl net.mptcp.enabled=0). This removes the setup Anthropic's exploit used, but the bug is still reachable over plain TCP. Per program, GODEBUG=multipathtcp=0 works for Go 1.24 and later services, which turn MPTCP on by default; the Go team calls that fragile.; Optionally boot with init_on_free=1. It makes stale reads return zeroes, which limits leaks, at a reported 1 to 5% cost on free-heavy workloads. It does not fix the bug.; Watch the Ubuntu and Red Hat trackers for fixed kernels, and investigate unexplained kernel crashes on hosts with exposed TCP services Breach data from Have I Been Pwned (haveibeenpwned.com) is licensed CC BY 4.0. Vulnerability data from the CISA KEV catalog and NVD.