CTIPilot
AI-generated · no human review · verify critical claims against the linked source. how it works →
‹Fri · 20 Nov 2026
All daily briefs →
Daily brief · UTC day

Friday, 20 November 2026

5 verified findings from 1 run · 6 updates to prior coverage · the settled record for this UTC day, in the classic brief order.

Criticality
Kind
Topic
Region
TL;DR · the day in one read
  1. 01Ruby 4.0.10, 3.4.13, 3.3.14 fix seven memory flaws; a Rails exploit is announced for 25 November: update the interpreter. The Ruby project released 4.0.10, 3.4.13 and 3.3.14 on 2026-11-18 for seven memory-corruption flaws in the regexp engine and in string case conversion, three rated Critical, and says it knows of proofs of concept that reach remote code execution. Veria Labs, which found the case-conversion flaw independently (CVE-2026-132825, a one-byte heap overflow), says Rails and classic-style Sinatra are exposed without authentication in their default configuration and plans to publish a code-execution proof of concept on stock Rails on 25 November 2026; no exploitation is reported. Ruby 3.2 and earlier are end of life, and the interpreter itself, not a gem, must be updated. →
  2. 02vCenter auth bypass exploited since September (GTIG), no workaround: update to 8.0 U3m, 9.0.2.0200 or 9.1.1.0200. Broadcom's advisory VMSA-2026-0009 of 2026-11-19 fixes CVE-2026-123890, an authentication bypass in the VMware Directory Service of vCenter that needs only network access to vCenter (CVSS 3.1 9.8), with no workaround and, in Broadcom's words, an emergency change. Google Threat Intelligence Group and Mandiant say ARMORY CASTLE, a suspected PRC-nexus actor, exploited it from early September 2026 to move from a compromised remote-support appliance to vCenter, forge administrator SAML sessions and extract domain-controller data from guest virtual disks on ESXi; Broadcom's advisory does not state exploitation. Fixed builds are 8.0 U3m, 9.0.2.0200, 9.1.1.0200 and VMware Cloud Foundation 5.2.5, and vCenter 7.0 has no public fix. →
  3. 03Mitel OpenScape 4000: two unauthenticated CVSS 10.0 Platform flaws; upgrade to V11 R2.14 or restrict access. Mitel's advisory MISA-2026-0013 of 2026-11-18 covers three flaws in the Platform component of OpenScape 4000 and OpenScape 4000 Manager: an unauthenticated improper-authentication flaw and an unauthenticated command injection, both CVSS 3.1 10.0, and a command injection reachable from the adjacent network (9.6). CVE ids have been requested but not assigned; the fix is V11 R2.14, PLT HF5 on V11 R1.26 or a patch for V10 R0.28 to V11 R0.22, and enabling Security Mode restricted access is Mitel's workaround. The advisory says nothing about exploitation, and it is the second critical Mitel advisory in a week. →

01Active threats, incidents & disclosures1 item

ROUTINEupdatedNATOB2

Università della Svizzera italiana (USI, Lugano): third parties exploited a vulnerability in an online-forms system and had access to form responses and attached documents from 9 to 17 November 2026, mainly contact details, and more than 8,000 e-mails warned those concerned; USI names no product or actor

USI said on 2026-11-19 that unauthorised third parties had access to data held on one of its applications between 2026-11-09 and 2026-11-17, by exploiting a vulnerability in the software of an online-forms system that the university uses to collect certain job applications, scholarship requests and surveys (USI via Corriere del Ticino, 2026-11-19). The affected server held the responses to the forms and their attached documents; USI's IT services secured it, USI informed the competent authorities and the people whose data may be affected, and it says the causes and the risk to those concerned are still being checked, which will take several more weeks (USI via laRegione, 2026-11-19). It names no product, vulnerability or attacker, and Ticinonline adds that it is not yet clear when the intrusion began or how long it lasted (Ticinonline, 2026-11-19).

Between 9 and 17 November 2026 unauthorised third parties had access to the data, exploiting a vulnerability in the software of an online-forms system used to collect certain types of job applications, scholarship requests and surveys (translated from Italian).

Corriere del Ticino (reproduces the USI communiqué) 2026-11-19

The data stolen are essentially first names, surnames, e-mail addresses and phone numbers, and to a minimal extent also job applications or scholarship requests (translated from Italian).

We sent more than 8000 e-mails to warn the people concerned by the data theft (translated from Italian).

Corriere del Ticino (interview with the USI chief operating officer) 2026-11-20
Updaterun 2026-11-21T0355Z-inteltitleheadlinesummarysourcesevidencesourcing_notebody

In an interview of 2026-11-20 USI's chief operating officer said the stolen data are essentially names, e-mail addresses and phone numbers and only to a minimal extent job applications or scholarship requests, that the forms software also serves event management, and that more than 8,000 e-mails warned those concerned (translated from Italian) (Corriere del Ticino, 2026-11-20). He calls the attack the work of AI agents that scan the network for vulnerabilities, offers no technical evidence for that, and leaves the product and the vulnerability unnamed (translated from Italian) (Corriere del Ticino, 2026-11-20).

incident20 Nov 05:09Zsingle-source · victim disclosureOpen finding →
HIGHCVE-2026-132825NATOA2

Ruby 4.0.10, 3.4.13 and 3.3.14: seven memory-safety flaws in the regexp engine and string case conversion, three rated Critical by the project, with proofs of concept for code execution; Veria Labs says Rails and classic Sinatra are exposed by default and announces a public Rails exploit for 25 November 2026 (CVE-2026-132825, CVSS 3.1 7.4)

The Ruby project released 4.0.10, 3.4.13 and 3.3.14 on 2026-11-18 to fix seven vulnerabilities in the regexp engine and in string case conversion, and says it is aware of proofs of concept that prove some of them can be used to trigger remote code execution (Ruby project, 2026-11-18). NCSC.ch lists three as Critical, a one-byte heap overflow in case conversion of strings in single-byte encodings such as ISO-8859-1 and Windows-1252, a double free in regexp compilation and a heap overflow from an integer overflow in regexp compilation, and the rest as High or Moderate (NCSC.ch, 2026-11-19); the project's advisory for the first confirms it can write one byte past the output buffer (Ruby project, 2026-11-18). Most regexp flaws need an application that compiles, matches or stringifies an attacker-controlled pattern, and one needs only a crafted subject string against a fixed case-insensitive backreference pattern; the case-conversion flaw needs attacker-controlled text passed to a case-changing method (NCSC.ch, 2026-11-19). Affected are Ruby 3.3.13 and earlier, 3.4.0 to 3.4.12 and 4.0.0 to 4.0.9, and Ruby 3.2 and earlier are end of life and unfixed (Ruby project, 2026-11-18).

NCSC.ch and watchTowr say Rails and classic-style Sinatra are exposed to the case-conversion overflow in their default configuration, and watchTowr reports no exploitation as of 2026-11-19 (NCSC.ch, 2026-11-19; watchTowr, 2026-11-19). Veria Labs, which says it found the flaw independently and first demonstrated code execution, calls it Ruby2Shell, says it is reachable without authentication on Rails and classic Sinatra and that plain Rack and other frameworks are not exposed by default, and plans to publish a proof of concept for remote code execution on stock Rails on 25 November 2026, or later if major vendors have not shipped a fix (Veria Labs, 2026-11-20). Veria lists GitHub Enterprise Server 3.22.5 and earlier and Chatwoot 4.20.0 and earlier as having no fixed release, and GitLab (fixed in 19.6.1, 19.5.8 and 19.4.10), Mastodon, OpenProject and the Discourse base image as patched; the Ruby advisories name no product, so the vendors' own advisories decide for each (Veria Labs, 2026-11-20). CVE-2026-132825 names the case-conversion overflow and is scored CVSS 3.1 7.4 with a local attack vector and high complexity (ENISA EUVD, 2026-11-19); the Ruby announcement says CVE identifiers were requested and none was assigned at publication, so the other six flaws are known by GHSA id (Ruby project, 2026-11-18).

We are aware of PoCs that prove some of them can be used to trigger remote code execution.

Ruby project 2026-11-18

We'll be publishing a proof of concept demonstrating remote code execution on stock Rails on 25 November 2026

Veria Labs 2026-11-20

Proof of Concepts will become available in a week.

NCSC.ch Cyber Security Hub 2026-11-19
vulnerability20 Nov 05:07Zmulti-sourceOpen finding →
HIGHCVE-2026-123890 +4exploitedNATOA2

CVE-2026-123890, VMware vCenter Server: an unauthenticated authentication bypass in the Directory Service gives an administrator session; GTIG reports the suspected China-nexus actor ARMORY CASTLE exploited it from early September, Broadcom lists no workaround; fixed in 8.0 U3m, 9.0.2.0200 and 9.1.1.0200 (CVSS 3.1 9.8)

Broadcom's advisory VMSA-2026-0009 of 2026-11-19 fixes CVE-2026-123890, an authentication bypass in the VMware Directory Service of vCenter: an actor with network access can bypass authentication and gain unauthorized access without credentials (CVSS 3.1 9.8) (Broadcom, 2026-11-19). Broadcom lists no workaround, calls the issue an emergency change and does not state exploitation (Broadcom, 2026-11-19). Broadcom credits Mandiant with the report (Broadcom, 2026-11-19), and Google Threat Intelligence Group and Mandiant say ARMORY CASTLE, a suspected PRC-nexus actor, exploited it in early September 2026 with its BONESAW utility, according to forensic evidence (GTIG, 2026-11-19). GTIG describes one intrusion that reached an internal vCenter from a Remote Support appliance compromised through the earlier BeyondTrust zero-day and names no victim sector or country (GTIG, 2026-11-19); BleepingComputer notes that Broadcom has not confirmed the exploitation (BleepingComputer, 2026-11-19).

The flaw is a session-key disclosure over DCE/RPC in the rpc_srp_verifier service on TCP 2012 that yields an authenticated administrator@vsphere.local session; BONESAW then read the Directory Service database, extracted the SAML signing key and created an SSO administrator, SAWSNAKE forged administrator SAML assertions, and the actor switched SSH on through the appliance management API, took the master symmetric key and the PostgreSQL credential file, switched SSH off again, logged in to ESXi hosts, planted an authorized_keys entry and extracted the domain database, registry hives and DNS data from Windows guest virtual disks before copying an archive out over SCP (GTIG, 2026-11-19).

Broadcom's response matrix names the fixes: vCenter 9.1 in 9.1.1.0200, 9.0 in 9.0.2.0200, 8.0 in 8.0 U3m, VMware Cloud Foundation 5.x in 5.2.5 through KB88287 and Telco Cloud through KB463116; vCenter 7.0 has no public fix, and extended-support customers contact Broadcom Support (Broadcom, 2026-11-19). The 8.0 U3m and 9.0.2.0200 updates block later upgrades to VCF 9.1.1 until 9.1.2 (Broadcom, 2026-11-19).

Triage: SSH enablement through the management API alone does not identify this activity; in GTIG's case it followed internal port sweeps from a non-management appliance and was reversed after the credential reads (GTIG, 2026-11-19).

ARMORY CASTLE exploited this vulnerability in early September 2026 using BONESAW, according to forensic evidence.

Google Threat Intelligence Group / Mandiant 2026-11-19

A malicious actor with network access to vCenter may exploit this issue to bypass authentication and gain unauthorized access to the system.

Broadcom (VMSA-2026-0009) 2026-11-19

The vCenter issue qualifies under ITIL methodologies as an emergency change, requiring prompt action from your organization.

Broadcom (VMSA-2026-0009 Q&A) 2026-11-19

While Broadcom has not confirmed that the flaw was exploited, researchers at Google's Mandiant have documented attacks exploiting CVE-2026-123890.

BleepingComputer 2026-11-19

CVE-2026-123890 VMWare vCenter Authentication Bypass Using an Alternate Path or Channel Vulnerability

CISA 2026-11-20

Builds on: BeyondTrust Remote Support zero-day: Rapid7 reports a Metasploit module; the actor's… · VMware vCenter's pre-auth Syslog RCE is exploited; patch it with the auth bypass and VMXNET3…

vulnerability20 Nov 05:02Zmulti-sourceOpen finding →
NOTABLENATOA2

Mitel OpenScape 4000 and OpenScape 4000 Manager, MISA-2026-0013: two unauthenticated flaws in the Platform component rated CVSS 3.1 10.0 (improper authentication, command injection) and an adjacent-network command injection (9.6); CVE ids pending, fixed in V11 R2.14, Security Mode restricted access is the workaround, no exploitation stated

Mitel's advisory MISA-2026-0013 of 2026-11-18 covers three flaws in the Platform component of OpenScape 4000 and OpenScape 4000 Manager, tracked by Mitel's internal ids because the CVE identifiers it requested are not yet assigned (Mitel, 2026-11-18). MTLVULN-1785 is an unauthenticated improper-authentication flaw that gives unauthorized access and MTLVULN-1786 an unauthenticated command injection through insufficient parameter sanitization, both CVSS 3.1 10.0 with changed scope; MTLVULN-1793 is a command injection that an unauthenticated attacker with adjacent network access can use, scored 9.6 (Mitel, 2026-11-18). Affected are OpenScape 4000 V11 R0.22 through V11 R2.11 and V10 R1.68 and earlier, and OpenScape 4000 Manager V11 R0.22 through V11 R2.11 and V10 R1.34, R1.42, R1.54 and R1.68; Mitel does not assess end-of-support releases (Mitel, 2026-11-18).

The issues are corrected in V11 R2.14, Mitel provides PLT HF5 for V11 R1.26 and a patch for V10 R0.28 to V11 R0.22 (for the Manager, V10 R1.34 to V11 R0.22), and for customers who cannot apply the fix quickly, enabling Security Mode restricted access is a workaround that prevents exploitation, with the steps in a portal-only article (Mitel, 2026-11-18). CERT-FR lists the risks as remote arbitrary code execution, remote denial of service, policy bypass and loss of integrity and confidentiality, and BSI describes an attacker who can bypass safeguards, disclose information or run arbitrary commands in the context of the system (CERT-FR, 2026-11-19; BSI, 2026-11-19). Mitel's text does not mention exploitation.

For customers who are not able to implement the recommended solution in a timely manner, enabling Security Mode restricted access can be used as a workaround to prevent exploitation of these vulnerabilities.

These issues are corrected in OpenScape 4000 V11 R2.14. Also, Mitel has provided PLT HF5 for version OpenScape 4000 V11 R1.26, and patch for releases V10 R0.28 to V11 R0.22.

Mitel (MISA-2026-0013) 2026-11-18

Builds on: Mitel MiCollab: unauthenticated CVSS 10.0 command injection in the conferencing component…

vulnerability20 Nov 05:08Zsingle-sourceOpen finding →
NOTABLECVE-2026-129289 +7NATOA2

PHP 8.5.13, 8.4.28, 8.3.37 and 8.2.36: CVE-2026-129289 lets an unauthenticated request smuggle a second FastCGI request into PHP-FPM and run PHP code when the front end reuses FastCGI connections; the release also fixes the session flaw left open in October and follow-ups to an incomplete earlier fix (CVSS 4.0 9.2)

The PHP project released 8.5.13, 8.4.28, 8.3.37, 8.2.36 and 8.6.0 on 2026-11-19, and the changelogs list up to 18 CVEs per branch (PHP, 2026-11-19). The most direct is CVE-2026-129289: when php://input was opened but not read to the end, PHP-FPM leaves the unread request bytes on a persistent FastCGI connection, so a malformed multipart body can place a second FastCGI request inside the first, and an unauthenticated attacker can bypass front-end routing and execute attacker-controlled PHP code with the privileges of the worker (PHP, 2026-11-19). It requires a front end that enables persistent backend connections and an application path that opens php://input without reading it fully; standard nginx and Apache configurations that close the FastCGI connection after each request are not affected by the demonstrated attack (PHP, 2026-11-19).

Two session flaws close the item the October release left open (watchTowr, 2026-11-20): CVE-2026-115560, where upload-progress session data is written in one serializer format and read in another, and CVE-2026-126714, where an interrupted write by the default files handler leaves a mixed old and new session file; both end in object injection and need an object gadget chain in the application (PHP, 2026-11-19; PHP, 2026-11-19). The fix for the first disables upload progress by default (PHP, 2026-11-19). CVE-2026-129481, a use-after-free of a stale session filename after an interrupted session shutdown, is in PHP's words an incomplete fix for CVE-2026-109619 and affects PHP 8.3 and later, and CVE-2026-129182, for PHP 8.5 only, covers further scenarios after the 8.5.12 fix (PHP, 2026-11-19; PHP, 2026-11-19).

CVE-2026-129474, PHP 8.5 only, is exposed on the built-in defaults, because fatal error backtraces are on in 8.5 and the default includes arguments, and php.ini-production, which sets zend.exception_ignore_args=1, or fatal_error_backtraces=0 prevents the attack; PHP confirmed code execution on 8.5 (PHP, 2026-11-19). CVE-2026-129496 needs a startup diagnostic written through a file-based error_log in a non-thread-safe build (PHP, 2026-11-19). CVE-2026-129291, an epoll use-after-free in PHP-FPM, gives root from a worker where catch_workers_output is on and the FPM master runs as root (PHP, 2026-11-19). NCSC.ch lists exploitation as unknown with no public proof of concept referenced, and watchTowr reports none as of 2026-11-20 (NCSC.ch, 2026-11-19; watchTowr, 2026-11-20).

Builds on: PHP2Shell: CISA lists both flaws as exploited after honeypot code execution; update the…

vulnerability20 Nov 05:04Zmulti-sourceOpen finding →

03Updates to prior coverage6 items

HIGHCVE-2026-102489 +1exploitedupdatedNATOA2

CVE-2026-102489 / CVE-2026-102490, Zammad helpdesk: a session-hijack remote code execution and a zammad-to-root escalation, both reported exploited since 21 September; 7.2.2 fixes the root flaw on DEB and RPM installs

First published 2026-10-02 · open finding →

Updaterun 2026-11-20T0356Z-inteltechniquessourcesbodycvesactionsheadlinesummary

DIVD completed its forensic investigation: the attacker, who used AI agents after reaching root, was mainly after credentials (a GitLab token, a Slack key and shared-mailbox OAuth tokens); about 41,000 ticket titles and a limited set of older conversations left the host, and the first response started 19 hours after the alert. Zammad's release 7.2.3 of 2026-11-11, a critical security update, supersedes 7.2.2 as the build to install.

DIVD completed its forensic investigation and updated its case file on 2026-11-18 (DIVD, 2026-11-18). It says that after obtaining root the attacker used AI agents to continue the attack, which it infers from the volume of payloads and the comments in the attack scripts (DIVD, 2026-11-18), and that the attacker was mainly interested in credentials, all of which have been rotated (DIVD, 2026-11-18). The credential material it touched was a GitLab personal access token that could not reach the internal GitLab, a Slack API key without read permission and the OAuth tokens of two shared mailboxes, used to send test emails with no mailbox-content compromise observed; only the host carrying Zammad was compromised at root level, and because direct outbound traffic from it was impossible the extraction from the ticket system was selective, about 41,000 ticket titles and a limited set of older conversations (DIVD, 2026-11-18); DIVD says its Jira and Confluence environment was also compromised, with 207 volunteer email addresses, two phone numbers and seven profile pictures taken (DIVD, 2026-11-18). DIVD says its alert arrived quickly but it took 19 hours to initiate the first response, and that there is plenty of room for improvement (DIVD, 2026-11-18). Separately, Zammad released 7.2.3 on 2026-11-11 as a critical security update that addresses several vulnerabilities, so 7.2.2 is no longer the build to install, and it requires an nginx or Apache2 configuration change that clears the Remote-User header next to X-Forwarded-User, which the Docker images apply automatically (Zammad, 2026-11-11).

NOTABLEupdatedNATOB2

ARTEX: Microsoft reports Storm-0940, a China-state aligned actor, running an open-source multi-agent LLM penetration-testing framework against hundreds of organizations, with privileged cloud access in under three hours in some cases, and Expel sees the same infrastructure abuse Okta self-service registration

First published 2026-10-22 · open finding →

Updaterun 2026-11-20T0356Z-intelsourcesbody

The European Commission confirmed an AI-generated cyber campaign against its internal systems, which it says it detected on 2026-10-22, with its HR and medical-reimbursement systems disrupted for more than 10 days per POLITICO; only an anonymous EU official names ARTEX, POLITICO places CERT-EU's finding of ARTEX use at a different date, and no access vector is stated.

The European Commission confirmed on 2026-11-19 that an AI-generated cyber campaign hit its internal systems: its spokesperson said the incident was detected on 2026-10-22 and confirmed on 2026-10-26, that several applications were taken offline as a protective measure and most are restored, that there is no evidence of a data breach affecting Commission staff as a whole and that it is not aware of a ransom request, and the Commission names no tool or attacker and declines to attribute or size the campaign (European Commission, 2026-11-19). POLITICO, citing an anonymous EU official, reports that CERT-EU detected earlier this month that international hackers used ARTEX, that the staff HR platform Sysper and the medical-reimbursement system MyPMO were disrupted for more than 10 days, and that investigators have not yet established whether staff data was compromised (POLITICO, 2026-11-19). The dates differ: the Commission says detection was on 2026-10-22, while POLITICO's official places CERT-EU's finding of ARTEX use earlier in November, and the sources do not reconcile them. The Commission has not confirmed the ARTEX link (European Commission, 2026-11-19), and the reporting does not establish whether ARTEX provided entry or was only used to find weaknesses (Cyber Security News, 2026-11-19).

NOTABLEupdatedNATOB2

Turla's Cereal toolset turns a compromised on-premises Exchange server into the command channel and mail collector at Kazakh and Romanian government bodies, with no attacker infrastructure and no new transport agent registered (ESET)

First published 2026-11-11 · open finding →

Improvementrun 2026-11-20T0356Z-intelsourcesbody

ReversingLabs' follow-up adds hunting specifics to Cereal, now in the Detection line: the loaders appear under other vendor names (BlankPage, CerealMood), a check for live HTTP.sys listeners, the proxy-DLL names to look for in Exchange and Windows service directories and module loads in WmiPrvSE.exe; it adds no victim or entry vector.

ReversingLabs correlated ESET's published loader samples and says the same files carry different vendor names, CerealLoader at ESET, BlankPage in Microsoft Defender and CerealMood in other engines, so a hunt on a single family name can miss matches; it says the work extends the artifacts defenders can search for and establishes no additional victims or initial-access method (ReversingLabs, 2026-11-19). Its hunting guidance (HTTP.sys listeners, proxy-DLL loaders such as version.dll and WTSAPI32.dll, and WmiPrvSE.exe module loads) is in the Detection line (ReversingLabs, 2026-11-19).

HIGHCVE-2026-114938updatedNATOA2

Java4Shell, CVE-2026-114938, OpenJDK 17.0.6 and later and Oracle Java SE on x86-64 CPUs with AVX-512 VBMI: a heap overflow in the Base64 decoder that its discoverer chains into pre-authentication code execution in Keycloak, Jira, Confluence, Artifactory, Jenkins and PingFederate; fixed in 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1, GreyNoise reports exploitation attempts, none confirmed successful against its sensors (CVSS 3.1 8.1)

First published 2026-11-18 · open finding →

Updaterun 2026-11-20T0356Z-intelentitiesaffected_productssourcesbodysummarycves

Elastic confirms Elasticsearch 7.17.3 and later is affected, with remote code execution needing an authenticated read-level user and unauthenticated actors limited to denial of service, and announces fixes without dates; Broadcom lists released Tanzu fixes and Palo Alto Networks lists fixed builds, none reporting exploitation.

Three more vendors published product status. Elastic says Elasticsearch 7.17.3 and later is affected with mitigating factors: remote code execution needs an authenticated user with at least read access on an affected CPU architecture, while unauthenticated actors can cause denial of service; Logstash 8.6.2 and later is affected only through the elastic_integration filter when an ingest pipeline decodes Base64 from an untrusted source, with a crash as the result, and Kibana, Beats, Elastic Agent, APM Server, the ECK operator and Elastic Maps Server are not affected, while Elasticsearch clusters that ECK manages follow the Elasticsearch rule (Elastic, 2026-11-19). Elastic plans fixes in 8.19.26, 9.4.10, 9.5.7 and 9.6.0 with a bundled JDK 27.0.1.1+4 and gives no dates, hot-fixed Elastic Cloud Serverless on 2026-11-18, and says fewer than about 5% of Elastic Cloud Hosted deployments, on Azure only, run an affected CPU architecture; a self-managed node is affected only on an x86-64 CPU with AVX512_VBMI (Elastic, 2026-11-19).

Broadcom's Tanzu advisory TNZ-2026-1107 says several Tanzu products ship affected Java runtimes and lists released fixes: Java Buildpack 5.5.0 and 5.5.1, Ruby Buildpack 1.10.104, Tanzu OpenJDK 17.0.21.1+5, 21.0.13.1+5 and 25.0.5.1+5 and Foundation Core 3.1.19, 3.3.10 and 11.0.3, where the identity service and CredHub carry an affected Java 25 runtime that Tanzu has not yet confirmed from the bill of materials; it names -XX:-UseBASE64Intrinsics as the preferred workaround (Broadcom, 2026-11-19). Palo Alto Networks lists fixed builds of Privilege Threat Analytics, Remote Access Connector and the Secure Access HTML5 Gateway, says PAN-OS is not affected, offers no workaround and is not aware of any malicious exploitation (Palo Alto Networks, 2026-11-20). None of these statuses changes the exploitation picture above.

NOTABLECVE-2026-115731 +1exploitedupdatedNATOA2

CVE-2026-115731, Puppet Enterprise before 2023.8.13, 2025.12.1 and 2026.0.1: deserialization of a crafted agent report gives code execution as the Puppet service user to any holder of a CA-signed agent certificate, and unauthenticated where the CA autosigns; CISA lists it as exploited (CVSS 4.0 9.4)

First published 2026-11-18 · open finding →

Improvementrun 2026-11-20T0356Z-intelsummaryaffected_productscvessourcesactionsbodyentities

The same advisory fixed CVE-2026-115732 (CVSS 4.0 9.0), a certificate-name authorization bypass that lets a certificate for a name extending a real host name authenticate as that host; the fixed builds are the same and Puppet Core 9.1.3 and 8.22.2 are added. CISA lists only CVE-2026-115731, and exploitation of the second flaw is not established.

The same Perforce advisory fixed a second flaw, CVE-2026-115732, an authorization bypass in the certificate-name matching of Puppet's authorization library, scored CVSS 4.0 9.0: a CA-signed certificate whose name extends or shortens an authorised host name is authorised as that host (ENISA EUVD, 2026-11-03). Neodyme's example is that an attacker who got a certificate for evil.target.lab.local can authenticate as target.lab.local, then fetch that host's catalog, which can carry configuration files with database credentials and API tokens, push spoofed facts and add exported resources (Neodyme, 2026-11-03). Where the CA autosigns, such a certificate can be requested for a name that does not exist yet, and OpenVox says autosign is not the default but is common in real deployments (OpenVox, 2026-11-03); NCSC.ch adds that in Puppet Enterprise the bypass also enables CA administration and remote command execution on managed nodes (NCSC.ch, 2026-11-19).

The fix is in the same Puppet Enterprise builds, in Puppet Core 9.1.3 and 8.22.2 and, for OpenVox Server and OpenVoxDB, in 9.2.0 and 8.18.0 (SecurityWeek, 2026-11-18; OpenVox, 2026-11-03). CISA's catalog entry covers only CVE-2026-115731 (CISA, 2026-11-17); NCSC.ch labels the pair actively exploited with a proof of concept available and lists a community report as an additional source (NCSC.ch, 2026-11-19), while the report The Hacker News quotes describes a CVE-2026-115731 intrusion, so exploitation of CVE-2026-115732 itself is not established (The Hacker News, 2026-11-18). NCSC.ch's temporary measures are to disable certificate autosigning, enforce strong request validation, audit the issued certificates for colliding names and review the CA configuration for signs of compromise (NCSC.ch, 2026-11-19).

NOTABLECVE-2026-133323 +4updatedNATOA2

Mitel MiCollab 10.3 (10.3.0.18) and earlier, MISA-2026-0012: an unauthenticated blind command injection in the Audio, Web and Video Conferencing component (CVSS 3.1 10.0) and four more flaws there, one an XML injection that can run commands (9.0); CVE-2026-133323 and CVE-2026-133303 lead the five assigned CVE ids, patch KB000129195, no exploitation stated

First published 2026-11-18 · open finding →

Improvementrun 2026-11-20T0356Z-inteltitlesummarycvessourcesbodyentities

MITRE assigned CVE ids to all five flaws on 2026-11-19: CVE-2026-133323 for the CVSS 10.0 command injection and CVE-2026-133303 for the 9.0 XML injection, with CVE-2026-133275, CVE-2026-133311 and CVE-2026-133324 for the three medium flaws. Mitel's advisory page still lists only its own tracking ids.

MITRE published CVE records for all five flaws on 2026-11-19, which Mitel's advisory page does not yet list. MTLVULN-1864 is CVE-2026-133323, the blind command injection (CVSS 3.1 10.0) (ENISA EUVD, 2026-11-19); MTLVULN-1862 is CVE-2026-133303, the XML injection that leverages pre-staged content (9.0) (ENISA EUVD, 2026-11-19); MTLVULN-1861 is CVE-2026-133275, the file upload (5.3) (ENISA EUVD, 2026-11-19); MTLVULN-1863 is CVE-2026-133311, the unauthorized access that needs an identifier or pre-staged content (5.9) (ENISA EUVD, 2026-11-19); and MTLVULN-1877 is CVE-2026-133324, the authentication bypass (5.3) (ENISA EUVD, 2026-11-19). Each record's score and affected range match Mitel's table, so the patch and the access restriction above are unchanged.

04Action items8 items

Verification & coverage notes1 run

2026-11-20T0356Z-intel · Sonnet 5.5 · window 26 h · 5 entries published

Verification & coverage notes

Window 26 h (gap 24.0 h to 2026-11-19T0355Z-intel, class Standard). Five new entries and six changelog records, no deep dive, no critical. The will-publish set was 14 items, over the ~8 a main-agent deep read should carry, so two scoped cti-research follow-ups read the primaries and returned claim tables whose passages were grep -F checked against saved bodies: DR1 (five vulnerability items, 2026-11-20T04:40:37Z to 05:03:10Z) and DR2 (eight incident and update items, 04:40:37Z to 05:05:59Z); the main agent composed every entry and record from those tables and the findings files. Stop-hook commit nudges arrived mid-run and were not acted on; the commit follows the gate and the verifier loop.

  • KEV sweep: kev_window_diff.py found no CISA KEV addition with dateAdded inside the window (catalog 2026.11.18) and no ransomware-flag row; S1 read the feed too and found the same. No row needed a disposition.
  • Backlog: the PHP row is published (2026-11-20/php-8-5-13-release-fpm-request-smuggling-session-flaws, notable, ; the October pre-announcement entry's "left unfixed" statements for CVE-2026-115560 are now stale and its update was reverted, see below). The Basel-Landschaft row stays held: S2 and S4 found only a class-level statement (the chancellery says the attacks are intrusion attempts, not DDoS flooding) and no vector, behaviour or data taken, no leak-site claim and nothing from BACS; expiry 2026-12-03 unchanged.
  • Priority calls: the vCenter authentication bypass (CVE-2026-123890) is high, not critical: GTIG documents one targeted intrusion that reached vCenter from an internal foothold, Broadcom does not state exploitation, and there is no public proof of concept, scanning or KEV listing, so the critical bar (mass exploitation imminent or an ongoing campaign) is not met on today's facts. Ruby is high because the discoverer announces a public Rails exploit for 2026-11-25 and Rails is exposed by default; PHP is notable (most flaws need non-default conditions, none is exploited), and the October PHP pre-announcement entry was NOT changed (see below).
  • S1 corrections by DR1: Mitel OpenScape 4000 has a workaround (Security Mode restricted access); vCenter exploitation is stated by GTIG; Ruby gained a dated exploit announcement and a product list from Veria Labs (single source).
  • European Commission: S2 and S4 both returned a new-incident item; it went in as an update on the ARTEX entry because the incident gives no vector, no actor and no behaviour of its own (incident floor) and ARTEX is named only by an anonymous official.
  • borderline-drop: F5 BIG-IP K000164058, six unauthenticated TMM stack overflows (S1, composed as a notable entry, dropped after verifier iteration 1); each needs specific iRule commands or an APM policy, F5's bulletin and K-articles are unreadable, the per-flaw conditions rest on CVE Program records the entry cannot cite, and no exploitation is reported; the previous fire dropped the same family. Re-open if F5 or a researcher reports exploitation.
  • Publica (S2, S4): not recorded. The entry already says test or migration data was held at the supplier; Publica's 2026-11-19 statement of the same origin and the external-review mandate add governance only.
  • borderline-drop: WordPress 7.1.5, primary 33 h old, headline flaw needs an Administrator to open a crafted link (Wordfence), auto-updates deliver it, no exploitation
  • borderline-drop: Splunk SVD-2026-1103; management-plane bypass needs reach to splunkd 8089 and an empty pass4SymmKey, no exploitation, regular patch cycle
  • borderline-drop: Dell PowerStore DSA-2026-479, NAS interfaces normally internal, no exploitation, regular patch cycle
  • borderline-drop: IBM Guardium DP 12.2, internal collector interfaces, no exploitation
  • borderline-drop: JetBrains TeamCity CVE-2026-132891, trigger conditions unestablished (vendor page unreadable), single ENISA record, no exploitation
  • borderline-drop: WatchGuard Dimension CVE-2026-86130, log server, no exploitation
  • borderline-drop: cPanel WP Toolkit proxy, single BSI relay, no CVE or conditions, vendor page unreadable
  • borderline-drop: Frappe GHSA-vwqr-4mjv-ffrm; exploited ITW but needs an authenticated portal user, email spoofing and attachment read only, no constituency nexus
  • borderline-drop: BSI Lagebericht 2026 (annual report), Germany-specific statistics, no Swiss figures, awareness only
  • borderline-drop: Canton Zug M365 policy, relaxation of an internal rule, no obligation change for other defenders
  • borderline-drop: nginx2root Swiss exposure counts (Risikomonitor via Netzwoche), banner-based, no primary
  • borderline-drop: depthfirst Narrow Grants, Wide Engines, patched CVEs on regular cycle, MySQL needs high privileges, Oracle needs Create Session, pre-login chain not reproduced, hardening advice generic
  • borderline-drop: React2STEAL, single source, attacker pipeline on GitHub, no defender decision beyond known React2Shell patching
  • borderline-drop: Phantom53/HavoX, victims US individuals and non-constituency regions, ClickFix chain commonplace
  • borderline-drop: Check Point State of Ransomware Q3 2026, outside window, statistics, no takeaway
  • borderline-drop: N-able Mail Assure intrusion, no vector, no data, no Swiss link
  • borderline-drop: Transport for London photocard breach, UK, no vector, no Swiss link
  • Also dropped by S4 (16 borderline_drops lines in findings.S4.yaml): Infoblox (repeat), Softway Medical, CNIL SAN-2026-016, Neuried, Contabo, DomainFactory, Wellington IT, the NVIDIA/Gentlemen claim, Lithuanian customs, IIHF and Reka, each without a constituency nexus or a transferable behaviour.
  • Not returned by S3 but flagged as possibly missed by earlier fires: Kaspersky's Operation ShadowCloud and AuditTeam OpsAgent posts (2026-11-17/18) and Talos UAT-12643 (2026-11-18), all outside the window; left for the audit's re-sweep.
  • Single-source: USI Lugano (single-source-victim, the victim's communique read through two near-verbatim relays); Mitel OpenScape 4000 (vendor advisory; CERT-FR and BSI restate it).
  • Rotation: 100 standard or candidate sources were excluded from the slices because the previous two fires attempted them (runs.recent_attempts); S1 26, S2 19, S3 14 and S4 8 slice records were all attempted, and belden-psirt and sygnia were extra probes recorded as outside-slice checks, not as attempts.
  • Coverage gaps: baselland-ch (403 on every transport); usi-ch (403, relays used); f5-k-articles (JavaScript shell, CNA records used); perforce-portal (shell, EUVD and press used); belden-psirt (403, reader pool dead); lab52 (listing undated, feed resets); golem article bodies (consent wall, titles only); swisspost-cybersecurity (no dates); senthorus-ch (weekly roundups only).
  • Reader pool: the jina keys answered HTTP 402 almost throughout; no source needed it.
  • Ledger note: S3 rewrote url-liveness.tsv twice (removing lines it had logged in error) while other agents appended, and DR2 reports two portal.perforce.com lines that record a 200 for JavaScript shells; neither is cited.
  • Home-region watch item: Canton Basel-Landschaft (held, expiry 2026-12-03). Re-check search for the next fire: Kanton Basel-Landschaft Cyberangriff Verwaltung Server neu installiert plus the BACS and NCSC.ch hub pages; publish when a source names a vector, behaviour or data taken.
  • Verifier iteration 1 findings declined with reason: the European Commission record stays an update on the ARTEX entry (a new government victim is a development PD-8 names; the section now states the date disagreement); CVE-2026-115732 gets no techniques[] id because no source names a technique the existing mapping lacks; the old PHP entry's main analysis length is left for the audit's compaction pass (this fire's record replaced sentences and added none).
  • Dedup warnings (informational): the vCenter and PHP entries share entities with earlier entries by product; both are distinct findings (a new flaw and a new release) and declare the nearest predecessors in references[].
  • Reverted update: the record on 2026-10-21/php-critical-release-pre-announced-cve-2026-109618-109619 (an improvement that moved its update target to 8.5.13, 8.4.28, 8.3.37 and 8.2.36 and fixed its CVE-2026-115560 statements) was reverted with git checkout after verifier iteration 5: touching the 81 KB critical entry put the whole file in verification scope, and the passes kept finding vendor lines (Cisco, Fortinet, GreyNoise start date, two evidence quotes, a Siemens version) that its cited pages had changed since it was last written, defects that predate this fire. Left for the quality audit: that entry still says CVE-2026-115560 is unfixed and names 8.5.12, 8.4.27, 8.3.36 and 8.2.35 as the targets, while the new PHP entry states the fix; it also needs the VulnCheck note that its exploit covers PHP 7.4.10 to 7.4.33, 8.0.15 to 8.0.30 and 8.1.10 to 8.1.34.
  • Verifier iteration 5 findings declined: the ENISA record for CVE-2026-132825 re-read at the main agent lists three references (Hacker News, the Ruby announcement, the GHSA) and no GitHub reproduction repository, so the Ruby entry states no public proof of concept; Tenable's updated distribution status for the Java4Shell entry (Ubuntu, Red Hat, Debian fixes) belongs to the previous fire's dated section and is left to the audit.
  • Verification: six iterations, none CLEAN; the loop ended on the low-residual early exit after iteration 6 (truth 2, editorial 0, no F1 or F4): both findings were fixed after that pass (the Java4Shell cves[].fixed Cisco qualifier restored, the PHP "left open in October" clause cited to watchTowr) and have not had a further pass, so verification_residual_count is 2. Iteration 1 dropped the F5 BIG-IP entry; iteration 5 led to reverting the October PHP entry's update. The later passes read pages that had changed since the text was written (Zammad 7.2.3, GitLab's Ruby fix, Cisco's updated product table), which is why the loop did not converge on CLEAN.