10 verified findings from 1 run · 3 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
01TYPO3: backend user to installer to admin and RCE (SA-027 Critical); Form Framework flaw SA-029 exploited; update now. TYPO3 published ten core advisories on 2026-11-17 for 10.0.0 to 14.3.11, fixed in 14.3.12 LTS, 13.4.40 LTS, 12.4.52 ELTS, 11.5.57 ELTS and 10.4.63 ELTS. The one rated Critical (SA-027) lets a file named FIRST_INSTALL placed in the web root re-enable the installer without authentication, which creates an administrator and leads to remote code execution, and SA-026 lets any low-privileged backend user create such a file. TYPO3 states that SA-029, a Medium information disclosure in the Form Framework that can become cross-site scripting through uploaded HTML, is known to be actively exploited, naming no actor or victim. →
02Java4Shell: OpenJDK Base64 heap overflow chained to pre-auth RCE in Jira and Keycloak; update Java, attempts seen. Oracle's Critical Security Patch Update and the OpenJDK advisory of 2026-11-17 fix CVE-2026-114938, an out-of-bounds write in the AVX-512 Base64 decoder that Calif, its discoverer, scores CVSS 4.0 9.2 and says it turned into pre-authentication code execution in the default configurations of Keycloak, Jira, Confluence, Artifactory, Jenkins and PingFederate; Oracle scores it CVSS 3.1 8.1 with high attack complexity. Only x86-64 hosts with AVX-512 VBMI are affected, a trigger proof of concept is public, Calif withholds its product chains, a second researcher claims a working exploit with no release stated, GreyNoise reports exploitation attempts without a confirmed success against its sensors, and VulnCheck says a working exploit is available to its customers. NCSC.ch and BACS warn operators to update internet-reachable Java applications and identity services first, and Cisco lists 13 products as affected with no fixed release yet. →
03Microsoft's agent-framework on PyPI carried a Python-startup credential stealer in 1.26.1 and 1.26.2; pin 1.26.0. Socket and SafeDep report that versions 1.26.1 and 1.26.2 of Microsoft's official agent-framework metapackage were published to PyPI on 2026-11-17, about 33 minutes apart, with a .pth file that runs a credential stealer whenever Python starts, so a pip install in a CI job is enough. It collects environment secrets and credential files and requests Azure managed-identity, Azure DevOps and, in 1.26.2, GitHub Actions OIDC tokens. Both versions were removed from PyPI and 1.26.0 is the last clean release; how the attacker obtained publishing access is not known. →
Mandiant reports a series of linked intrusions in which COLD RELIC, the name Google Threat Intelligence Group (GTIG) now uses for the group formerly tracked as COLDRIVER and UNC4057 (also Star Blizzard, Callisto and Gossamer Bear), turned data stolen in one US organisation into the pretext for the next (GTIG, 2026-11-17). GTIG cites a US Department of Justice assessment that the group operates within Russia's FSB and names government officials, defense contractors and NGOs in Europe and North America among its targets (GTIG, 2026-11-17). In 2026 the operators approached a second US organisation over an encrypted messaging app while impersonating an employee of the first, shared phishing links that stole Entra ID credentials and tokens and sent the victim on to document lures matching the conversation, and registered a new MFA device on the account (GTIG, 2026-11-17).
In the late-2025 case the operators used the stolen credentials of a Super Admin account from a US residential proxy network close to the victim, pressed the user with MFA push notifications until one was accepted, registered a device model new to the tenant and kept access with passkeys, MFA backup codes and rogue MFA devices (GTIG, 2026-11-17). They took the whole cloud drive by mass sharing to an operator-run mailbox and by direct download, queried the admin investigation and email audit logs to check defensive visibility, viewed, modified and deleted Alert Center notifications, exported the employee list and ran a tenant-wide Customer Takeout export (GTIG, 2026-11-17). Access lasted several months through OAuth mail sync to an external client and re-authentication with enrolled passkeys that survived password and session resets, and in one case a password reset without revoking sessions did not evict the operators (GTIG, 2026-11-17). They searched for credentials, personnel data and .pst archives (GTIG, 2026-11-17).
Triage: a directory export, a Customer Takeout or a passkey or device enrolment by an administrator can be routine; the cues in GTIG's account are an unfamiliar device, a non-corporate or residential-proxy network, or an enrolment right after an MFA-fatigue pattern (GTIG, 2026-11-17).
COLD RELIC is a Russian state-sponsored threat group that the US Department of Justice (DOJ) has assessed operates within Russia's Federal Security Service (FSB).
COLD RELIC leveraged passkeys, MFA backup codes, and rogue MFA devices to retain access throughout the intrusion.
Reauthentication using unrevoked attacker-enrolled passkeys that persisted across password and session resets and bypassed device registration controls
Socket reports that the Microsoft agent-framework PyPI package was compromised on 2026-11-17, in versions 1.26.1 and 1.26.2, both of which install Python startup hooks that collect credentials, environment data and cloud identity tokens and send them to a web endpoint (Socket, 2026-11-17). SafeDep, which flagged both releases independently, says the project is Microsoft's official metapackage, published under the microsoft account on PyPI, that 1.26.1 appeared at 20:35 UTC and 1.26.2 at 21:08 UTC, that each is a single wheel where earlier releases shipped a wheel and a source archive, and that both were removed from PyPI and 1.26.0 is the last clean release (SafeDep, 2026-11-18). The attacker published 1.26.2 after 1.26.1 was gone, so removing one version did not stop it, and how it obtained publishing access is not known (SafeDep, 2026-11-18).
Nothing has to import the package: a .pth file runs on Python startup, imports a module and launches collection in a background process or thread, so a pip install in a CI job is enough (Socket, 2026-11-17; SafeDep, 2026-11-18). 1.26.1 routes through the package's own initializer, while 1.26.2 moves the payload to a new top-level module and restores the legitimate initializer byte for byte, so checking only the initializer misses it (Socket, 2026-11-17). The code matches environment variables by substring, such as TOKEN, SECRET, KEY and AZURE, reads other processes' environments and credential files for pip, npm, netrc, git, Docker, Hugging Face and Foundry, and actively requests fresh Azure managed-identity and Azure DevOps tokens, with GitHub Actions OIDC tokens added in 1.26.2; it compresses the result and posts it to a web endpoint and downloads no further code (Socket, 2026-11-17). There is no persistent on-disk marker, so fresh process trees can trigger collection again (Socket, 2026-11-17).
On Python startup, the .pth file imports a module that launches collection in a background process/thread.
RIKEN's Center for Computational Science says that between 2026-11-10 and 2026-11-13 an external third party entered Fugaku's internal operational network through the publicly accessible web page that displays the supercomputer's operating status, and that the exact method is still under investigation (RIKEN, 2026-11-17). The intruder tampered with multiple internal servers, some of them with malicious programs, and took or may have taken files in users' home directories and user-authentication data, including credentials users had stored for connecting to other systems, such as SSH private keys and access keys; up to 4,222 Fugaku users and 375 Rikyu users are affected (RIKEN, 2026-11-17). RIKEN has confirmed that the intruder tried some of the obtained credentials against other systems and is checking whether damage or secondary misuse followed (RIKEN, 2026-11-17).
RIKEN and the external provider that maintains the system began investigating suspicious network traffic on 2026-11-12, confirmed the access on 2026-11-13 and suspended Fugaku on 2026-11-14 (RIKEN, 2026-11-17). It replaced the administrative keys, invalidated the public keys all users use to log in and halted all running and queued jobs, and it names no actor; it says there is no evidence yet that research data was exfiltrated (RIKEN, 2026-11-17; BleepingComputer, 2026-11-17).
an external third party gained unauthorized access to Fugaku’s internal operational network by using a publicly accessible web page displaying Fugaku’s operational status (the dashboard).
The exact method that was used to gain unauthorized access is currently under investigation.
Furthermore, we have confirmed that the unauthorized party attempted to connect to other systems using some of the obtained credentials.
RIKEN
incident18 Nov 05:08Zsingle-source · victim disclosureOpen finding →
Barracuda says the attacker did not break in but used a valid cloud access key that Barracuda had issued to a third-party vendor, a company working in AI, for an earlier project and had not retired when the work ended (Barracuda Blog, 2026-11-11). The key had been stored in a configuration file in the vendor's GitHub repository, which was exposed when the vendor's development environment was compromised (Barracuda Trust Center, 2026-11-13). The first call with the key came on 2026-11-01 at 04:33 UTC; within minutes the attacker was mapping what the key reached and reading stored secrets such as AWS access keys and Azure service principal secrets, and each secret opened the next system, from cloud accounts to a data analytics system, to a source-code access token and on to Kubernetes clusters (Barracuda Blog, 2026-11-11). In just under four hours on 2026-11-02 it pushed short-lived branches to a handful of repositories to trigger a build workflow written to print the build system's secrets into the job log, encoded so that log masking would not hide them, and its last recorded activity was at 23:31 UTC, about 10 cloud accounts and 44 hours after the start (Barracuda Blog, 2026-11-11).
In a Kubernetes cluster the attacker deployed a container that mounted the underlying server's file system read-only, enough to harvest that server's credentials, and tried to create a new administrative cloud identity, which was blocked (Barracuda Blog, 2026-11-11). On servers where it had a shell, its commands searched for environment variables named WALLET, MNEMONIC, SEED and PRIV, Ethereum keystore files, extended private keys, seed phrase files and Kubernetes secrets with wallet-related names, and Barracuda calls the activity consistent with a financially motivated actor seeking cryptocurrency and reusable credentials (Barracuda Trust Center, 2026-11-13). Barracuda says the resource names, the sequence of activity and the credential and wallet hunting routines match ARTEX, which many groups use, so the match shows how the attacker operated, not who it was (Barracuda Blog, 2026-11-11).
The investigation has found no direct evidence that customer integration credentials were viewed or accessed.
The cloud access key used had been stored in a configuration file within the third party’s GitHub repository, which was exposed when the third party’s development environment was compromised.
Barracuda Trust Center
Because many groups use ARTEX, the match tells us how the attacker operated, not who it was.
Oracle's Critical Security Patch Update and the OpenJDK advisory of 2026-11-17 list CVE-2026-114938 in the hotspot compiler component, which Oracle rates CVSS 3.1 8.1 with high attack complexity (Oracle, 2026-11-17; OpenJDK, 2026-11-17). Calif, which found it, says HotSpot's x86-64 Base64 intrinsic, which uses AVX-512 VBMI instructions, calculates a negative output length for a short padded input and that length becomes a full 64-byte write that sends 62 bytes past the end of the array; only the landing position of the write is controllable, a misplaced write can crash the JVM (Calif, 2026-11-17). It scores the flaw CVSS 4.0 9.2 and reports pre-authentication code execution in the default configurations of Keycloak, Jira, Confluence, Artifactory, Jenkins and PingFederate, authenticated execution in several further server products, and a client-side path in which a Java program that opens a TLS connection to a malicious server is compromised (Calif, 2026-11-17). Tenable relays Immersive's claim of a working exploit with no release stated (Tenable, 2026-11-17).
The flaw needs an x86-64 processor with AVX-512 VBMI and affects OpenJDK from 17.0.6 on, ARM64 hosts are not affected (Calif, 2026-11-17). CISA published an alert to patch immediately (CISA, 2026-11-17), and the NCSC.ch Cyber Security Hub says the published research demonstrates exploitation without establishing malicious exploitation (NCSC.ch, 2026-11-17); Talos has seen no exploitation in the wild yet and expects probing (Talos, 2026-11-18), while Tenable relays GreyNoise's report of exploitation attempts as of 2026-11-18, none confirmed successful against its sensors, and VulnCheck's statement that a working RCE exploit is available to its customers (Tenable, 2026-11-17).
Our initial testing demonstrated remote code execution in the following widely deployed Java applications, in their default configs.
Only the landing position of the out-of-bounds write is controllable.
We are holding back the application-specific exploits for now.
Because the vulnerable code is hardware-specific, Java4Shell also requires an x86-64 processor with AVX-512 VBMI support. ARM64 CPUs are not affected.
Cisco published its product impact on 2026-11-18: Catalyst Center, Secure Network Analytics, Secure Workload and ten more products are affected, each with the fixed release still to be determined, Cisco offers no workaround, IOS XE and UCS Central are under investigation, and its PSIRT is aware of proof-of-concept code but not of malicious use (Cisco, 2026-11-18). Fortinet says it is investigating the impact on its products and lists none as affected (Fortinet, 2026-11-18). Tenable relays GreyNoise's report of exploitation attempts as of 2026-11-18, none confirmed successful against its sensors, and VulnCheck's statement that a working RCE exploit is available to its customers (Tenable, 2026-11-17). BACS's warning of 2026-11-18 asks operators to prioritise internet-reachable applications and central sign-in services, to include internal systems, to clarify other Java software such as banking software with their suppliers and to have externally operated services confirm the fix, and says an update removes no compromise that has already happened (translated from German) (BACS, 2026-11-18). Tenable adds that as of 2026-11-18 Red Hat and Ubuntu had released no fix and Debian had fixed only its unstable branch, so hosts on those distributions had no stable vendor build yet, and that the Keycloak project recommended starting its server with -XX:-UseBASE64Intrinsics as an interim mitigation, its x86-64 container images tagged 26.8.2-1 and 26.8.2-2 carrying an affected Java build (Tenable, 2026-11-17).
TYPO3 released ten core advisories on 2026-11-17, fixed in 14.3.12 LTS, 13.4.40 LTS, 12.4.52 ELTS, 11.5.57 ELTS and 10.4.63 ELTS (TYPO3, 2026-11-17); versions 10.0.0 to 10.4.62, 11.0.0 to 11.5.56, 12.0.0 to 12.4.51, 13.0.0 to 13.4.39 and 14.0.0 to 14.3.11 are affected (TYPO3 SA-027, 2026-11-17). SA-027 is rated Critical: placing a file whose name starts with FIRST_INSTALL in the public web root or the project root re-enables the installer on a configured instance without authentication, which allows changing the database connection and creating a new administrator and system maintainer account, and that leads to remote code execution; leftover or restored marker files also work (TYPO3 SA-027, 2026-11-17). SA-026, rated High, supplies the file: the backend endpoint for adding online media accepts a target folder in the fallback storage, which points to the public web root, so any authenticated low-privileged backend user can create media files with a partially controllable name there (TYPO3 SA-026, 2026-11-17).
TYPO3 states that SA-029, a Medium information disclosure in the Form Framework, is known to be actively exploited, and names no actor, victim or behaviour (TYPO3 SA-029, 2026-11-17). A failed upload shows the raw storage error with the upload sub-folder, so an attacker can derive the public URL of an uploaded file, and where the form accepts HTML this becomes cross-site scripting unless the web server applies TYPO3's recommended Content-Security-Policy to the upload folder (TYPO3 SA-029, 2026-11-17). SA-032 closes the matching gap, an empty allowedMimeTypes list that was treated as no restriction and let unauthenticated visitors upload HTML or SVG to a public folder (TYPO3 SA-032, 2026-11-17). SA-028 lets ImageMagick's MVG coder disclose any web-server-readable file through a processed image; it needs ImageMagick with its default configuration allowing the MVG coder, and unauthenticated visitors reach it where such an upload form exists (TYPO3 SA-028, 2026-11-17). SA-025 is an Import/Export path traversal that writes arbitrary files and needs an administrator or a user enabled for import (TYPO3 SA-025, 2026-11-17). Four lesser flaws are an incomplete open-redirect fix (SA-030), page-cache flooding where cHash validation is not enforced (SA-031), file-dump token reuse (SA-033) and an SSRF in the 14.x Dashboard RSS widget (SA-034) (TYPO3 SA-030, 2026-11-17; SA-031; SA-033; SA-034).
This vulnerability is known to be actively exploited.
The installer then became accessible without authentication, allowing for example to change the database connection and to create a new administrator and system maintainer account
any authenticated, low-privileged backend user, even without file mounts or module access, could create online media files with a partially controllable file name in the public web root.
Mitel's advisory MISA-2026-0012, first issued on 2026-11-16, covers five flaws in the Audio, Web and Video Conferencing (AWV) component of MiCollab 10.3 (10.3.0.18) and earlier; Mitel tracks them by internal ids because the CVE identifiers it requested are not yet assigned (Mitel, 2026-11-16). MTLVULN-1864 lets an unauthenticated attacker conduct a blind command injection through insufficient parameter sanitization and is scored CVSS 3.1 10.0 with changed scope; MTLVULN-1862 lets an unauthenticated attacker leverage pre-staged malicious content for an XML injection that could run arbitrary commands, scored 9.0 with high attack complexity (Mitel, 2026-11-16). Three further flaws are rated Medium: an unauthenticated arbitrary file upload through weak file-type validation (MTLVULN-1861, 5.3), unauthorized access that needs a system-generated random identifier or pre-staged content (MTLVULN-1863, 5.9) and an authentication bypass (MTLVULN-1877, 5.3) (Mitel, 2026-11-16). Mitel says exploiting them together can significantly amplify their impact (Mitel, 2026-11-16).
The fix is a patch for release 10.3 (10.3.0.18), delivered through knowledge-base article KB000129195 on Mitel's customer portal, or an upgrade to a later release that Mitel does not name; Mitel's partners can supply it to customers without portal access (Mitel, 2026-11-16). Mitel's July advisory named 10.3 (10.3.0.18) as the release that fixed its earlier AWV command injection, so a system moved there in July is inside the new affected range (Mitel, 2026-07-22). CERT-FR lists the advisory on 2026-11-17 and BSI rates it high, describing a remote anonymous attacker who can manipulate files, run code with the service's rights and bypass security measures (CERT-FR, 2026-11-17; BSI, 2026-11-17).
could allow an unauthenticated attacker to conduct a blind command injection attack due to insufficient parameter sanitization.
the risk may be mitigated by restricting external access to the AWV component.
Exploiting these vulnerabilities together can significantly amplify their impact.
Oracle's Critical Security Patch Update of 2026-11-17 carries 655 new patches (Oracle, 2026-11-17). Its risk matrices score 62 rows at CVSS 3.1 9.0 or above, 56 of them marked remotely exploitable without authentication and without privileges (Oracle, 2026-11-17); Tenable counts 62 critical patches across 57 CVEs (Tenable, 2026-11-17). Six of them score 10.0 with changed scope: Oracle Access Manager 12.2.1.4.0 and 14.1.2.1.0 in the Admin Console (CVE-2026-114835) and twice in the Authentication Engine (CVE-2026-91500, CVE-2026-91510), WebLogic Server TopLink Integration (CVE-2026-91569), Platform Security for Java (CVE-2026-114824) and Business Intelligence Enterprise Edition Platform Security (CVE-2026-114784), all over HTTP (Oracle, 2026-11-17).
WebLogic Server 12.2.1.4.0 to 15.1.1.0.0 carries further unauthenticated rows at 9.8: Core over T3 and IIOP (CVE-2026-91470) and over SOAP (CVE-2026-91456), and TopLink Integration over HTTP (CVE-2026-91570, CVE-2026-91583) and TCP (CVE-2026-91585); Oracle words the T3 and IIOP row as an easily exploitable flaw that lets an unauthenticated attacker with network access compromise WebLogic Server (Oracle, 2026-11-17). E-Business Suite 12.2.3 to 12.2.15 has four unauthenticated HTTP rows at 9.8 in Attachments (CVE-2026-114880), Application Object Library (CVE-2026-114924), Applications Framework REST Services (CVE-2026-114925) and Universal Work Queue (CVE-2026-114863), and APEX and REST Data Services have two at 9.8 over HTTPS (CVE-2026-114849, CVE-2026-114850) (Oracle, 2026-11-17). Oracle's footnote says the patch for the E-Business Suite row CVE-2026-114918 also addresses CVE-2025-61882 (Oracle, 2026-11-17), which CISA lists as exploited with known ransomware use (CISA, 2026-11-17). The release also carries three rows for the Brotli flaw CVE-2026-111497 in Communications Diameter Signaling Router, Communications Unified Assurance and PeopleSoft PeopleTools (Oracle, 2026-11-17).
Oracle's matrices carry no exploitation flag, and its interim relief is blocking the network protocols an attack requires or, for attacks that need certain privileges or package access, removing those; patches exist only for versions under Premier or Extended Support, and Oracle says earlier versions are likely affected too (Oracle, 2026-11-17).
This Critical Security Patch Update contains 655 new security patches across the product families listed below.
Easily exploitable vulnerability allows unauthenticated attacker with network access via T3, IIOP to compromise Oracle WebLogic Server.
Oracle
This month's update includes 62 critical patches across 57 CVEs.
Drupal shipped its announced core security release on 2026-11-18. SA-CORE-2026-017 (CVE-2026-132165, rated Critical 18 of 25, with access complexity Complex, access none, exploit status Proof, meaning a proof of concept exists with documented methods for developing an exploit in the wild, and target distribution Uncommon, meaning only uncommon module configurations are exploitable) says that under specific conditions an attacker can use the installer to attempt to reinstall Drupal, which can lead to a site takeover or remote code execution (Drupal Security Team, 2026-11-18). SA-CORE-2026-018 (CVE-2026-132166, Moderately critical 14 of 25) is a gadget chain that Drupal says is not directly exploitable: a separate vulnerability must be present to let an attacker pass unsafe input to unserialize() (Drupal Security Team, 2026-11-18).
Both advisories cover core before 10.6.22, from 11.0.0 before 11.3.22 and from 11.4.0 before 11.4.13, fixed in 11.4.13, 11.3.22 and 10.6.22; Drupal 11.2.x and below and 10.5.x and below are end-of-life and receive no security coverage (Drupal Security Team, 2026-11-18). The only workaround Drupal gives is to secure the installer; it adds that some exploit paths are mitigated if the web server cannot write to or change permissions on settings.php, and that sites on Drupal Steward are already protected from SA-CORE-2026-017's known attack vectors (Drupal Security Team, 2026-11-18). The pre-announcement had warned that exploits might be developed within hours or days (Drupal Security Team, 2026-11-16); NCSC.ch's Security Hub rated exploitation unknown on 2026-11-17 (NCSC.ch, 2026-11-17), and neither advisory reports exploitation.
The Drupal Security Team urges you to reserve time for core updates at that time because exploits might be developed within hours or days.
The vulnerabilities are currently rated as critical and at least one can be exploited by anonymous users.
Neither the Security Team nor any other party is able to release any more information about this vulnerability until the announcement is made.
Under specific conditions, an attacker can use the installer to attempt to reinstall Drupal. This can lead to a site takeover or remote code execution.
The release the pre-announcement described is out. Drupal published SA-CORE-2026-017 (CVE-2026-132165, Critical 18 of 25) and SA-CORE-2026-018 (CVE-2026-132166, Moderately critical 14 of 25, not directly exploitable) on 2026-11-18, fixed in Drupal 11.4.13, 11.3.22 and 10.6.22 (Drupal Security Team, 2026-11-18, Drupal Security Team, 2026-11-18). SA-CORE-2026-017 says an attacker can use the installer to attempt to reinstall Drupal, which can lead to a site takeover or remote code execution, and the only workaround is to secure the installer (Drupal Security Team, 2026-11-18). The 18 of 25 rating in the pre-announcement is SA-CORE-2026-017's; neither advisory reports exploitation.
CISA added CVE-2026-115731 to its Known Exploited Vulnerabilities catalog on 2026-11-17 on evidence of active exploitation, naming no actor or victim (CISA, 2026-11-17). The flaw sits in Puppet's report-processing code as used by Puppet Server: a client that holds any CA-signed agent certificate can run arbitrary code on the server as its service user by sending a report with a crafted rich-data type in a PUT request, and where the certificate authority autosigns requests without strong verification an unauthenticated network attacker can first obtain such a certificate (ENISA EUVD, 2026-11-03). The CNA scores it CVSS 4.0 9.4 (ENISA EUVD, 2026-11-03) and Red Hat CVSS 3.1 9.9 (Red Hat, 2026-11-03).
The discoverers at Neodyme call the bug authenticated in the strict sense, because it needs a valid agent certificate, and say it allows code execution on the management server in a default configuration (Neodyme, 2026-11-03). The OpenVox advisory of the same day includes a full trigger event and says autosign is common in enterprise deployments and that code execution as the puppet user gives access to the CA signing key (OpenVox, 2026-11-03). Puppet's 2026.0.1 release notes list the flaw among the fixes in Puppet Enterprise 2026.0.1 (Puppet, 2026-11-12), and the EUVD record names 2025.12.1 and 2023.8.13 as the fixed releases of the other two lines (ENISA EUVD, 2026-11-03); OpenVox fixes it in server 9.2.0 and 8.18.0 (oss-security, 2026-11-03), and Vox Pupuli says open-source Puppet will not receive any more updates (Vox Pupuli, 2026-11-03).
allows a client that holds any CA-signed agent certificate to execute arbitrary code on the Puppet Server as its service user by submitting a report containing a crafted rich-data type via PUT request.
Axis's advisory of 2026-11-16 says AXIS OS version 6.50.5.10 and higher is affected by CVE-2026-119618 (CVSS 9.8 per Axis), that it has no fixed release yet and plans a service release for all affected tracks within a couple of weeks, and gives a workaround that sets two HTTP/2 device parameters to no.
Axis published a security advisory on 2026-11-16 saying AXIS OS version 6.50.5.10 or higher is affected by CVE-2026-119618, which it describes as a CVSS 9.8 remote code execution flaw that would let an unauthenticated attacker gain administrator access to the device, and which it traces to the Apache HTTP Server release 2.4.71 that fixes it (Axis, 2026-11-16). Axis names no tracks and no fixed version: it is working on a service release for all affected AXIS OS tracks within the next couple of weeks, with more accurate timelines to follow, and devices outside those tracks that are still supported will be patched according to their maintenance schedule (Axis, 2026-11-16). To Axis' knowledge no exploit is public and it is not aware of exploitation in its network products (Axis, 2026-11-16).
As an immediate workaround Axis sets the device parameters System.HTTP2.AllowH2 and System.HTTP2.AllowH2c to no through param.cgi, to be reverted once an update is installed, and suggests Axis Device Manager or another configuration tool for mass deployment (Axis, 2026-11-16).
N-able released Hotfix 3 (build 2026.4.1.39, 2026-11-16): 17 more CVE ids fixed, Agent and Probe authentication hardened, the Legacy API permanently blocked, and every Agent and Probe must be upgraded even from Hotfix 2. N-able does not say Hotfix 3 fixes 4EverAble2Shell or Agent2Shell, and Huntress reports public proof-of-concept code for all 17 ids.
N-able released N-central 2026.4 Hotfix 3, build 2026.4.1.39, on 2026-11-16 as a critical security update that builds on and supersedes Hotfix 2; it fixes 17 further CVE ids, adds missing authentication and permission checks to server operations, hardens how Agents and Probes authenticate with the server, adds validation to file publishing, path and XML import handling, and permanently blocks the legacy APIs (N-able, 2026-11-16). Agents and Probes that are not upgraded are susceptible to impersonation, and all customers, hosted and on-premises, must upgrade every Agent and Probe to 2026.4.1 HF3 whether or not they already run Hotfix 2 (N-able, 2026-11-16).
The release notes name neither 4EverAble2Shell nor Agent2Shell nor the Apache Axis flaw and do not say they are fixed; what changes is the Legacy API, which N-able itself said both chains need and which is now blocked permanently (N-able, 2026-11-16; Huntress, 2026-11-17). Among the 17 ids, the release notes title CVE-2026-129272 "MMS3 appliance authentication bypass via predictable passcode", CVE-2026-129280 "ServerUI SOAP operations missing authorisation checks" and CVE-2026-129286 "SIS GenericFiles version path traversal leads to RCE", and Huntress counts three critical, eight high and six medium scores, all assigned by N-able as CVE numbering authority (N-able, 2026-11-16; Huntress, 2026-11-17).
Huntress reports that public proof-of-concept code for all 17 CVE ids appeared on the evening of 2026-11-16, shortly after the hotfix, says any unpatched server should be treated as exposed, and has not observed exploitation of them in its partners' environments (Huntress, 2026-11-17). Field Effect reports that technical details on 24 further N-central vulnerabilities were published and are reported to be addressed by Hotfix 3, a count that the 17 CVE ids in N-able's notes do not reconcile (Field Effect, 2026-11-09). Huntress advises confirming the Patch Management Engine is at 2.14.4.5829 or later, rotating secrets embedded in Automation Manager policies on a server that was exposed before patching, not relying on Take Control audit records from an unpatched server, and enforcing MFA on all N-central accounts (Huntress, 2026-11-17).
Optimizely's own status entry says an unauthorized party exploited a vulnerability in Experimentation to add the JavaScript to some customer projects, Report URI counts 178 organisations that served the code on 2026-11-16, and the Optimizely files carried the code from 2026-11-16 rather than 2026-11-13. The loader also hides in base64 where a host-name search misses it.
Optimizely says an unauthorized party exploited a vulnerability in Experimentation and used it to add malicious JavaScript to some customer projects, that it has fixed the vulnerability and removed the injected code, and that it secured affected user accounts; that post named no flaw class, CVE or count, and the status entry's later post of 2026-11-18 is described below (Optimizely, 2026-11-18). The status entry covers the Optimizely files only; how the Kissmetrics, StorifyMe and minute-ly scripts were reached is still unknown.
Report URI's timeline adds that the card skimmer was added to the clipper at 13:43 UTC on 2026-11-16, that a mass clean-up ran between about 20:40 and 21:10 UTC, and that a snippet cleaned and then re-infected shows the writer still had access at that point (Report URI, 2026-11-16). Contradiction on dates: Ohey puts the change to the Optimizely project files at 2026-11-13 together with the Kissmetrics and StorifyMe scripts (Ohey, 2026-11-17), while Report URI's first Optimizely injection is at 04:21 UTC on 2026-11-16; Report URI's logged first-hand record is the better dated one, and the earlier Kissmetrics, StorifyMe and minute-ly wave is not in the 178 (Report URI, 2026-11-16).
The injected loader came in three forms: a plain script tag, the same address hidden in base64 and decoded with atob(), which a search for the host name misses, and a full inline copy that runs through eval (Report URI, 2026-11-16). Cards entered in hosted payment iframes were out of the skimmer's reach, while cards typed into a site's own form fields were not (Report URI, 2026-11-16).
On every x86-64 host with an AVX-512 VBMI processor that runs Java 17, 21, 25 or 27, update the JDK to 17.0.21.1, 21.0.13.1, 25.0.5.1 or 27.0.1.1 (Oracle builds; other distributions use their own build numbers) or 8u513-perf, then restart the Java processes and rebuild containers, internet-reachable applications and the identity, build and artifact servers Calif names (Keycloak, Jira, Confluence, Artifactory, Jenkins, PingFederate) first; where that waits, start the JVM with -XX:-UseBASE64Intrinsics, or with -XX:UseAVX=2 on x86-64 hosts only.
List every page of your organisation's sites, and of the agencies that build them, that has loaded Optimizely project JavaScript since 2026-11-16, or Kissmetrics i.js, a StorifyMe widget or the minute-ly loader since 2026-11-13, capture the bytes the browser receives and compare them with the provider's own project content (a search for the loader's host name misses the variant that hides it in base64), and where a file carries the injected loader remove it, keep the response bytes as evidence and treat the payment, login and form pages that loaded an affected Optimizely file as exposed to card and session theft.
If your organisation or an agency building its sites has an Optimizely Web Experimentation account, audit every Project JavaScript field for code nobody wrote and rotate the API tokens, session tokens and integration keys tied to it, because the attacker impersonated users on targeted accounts to edit that field.
In every Google Workspace and Entra ID tenant you run, search the administrator and sign-in logs for passkey, backup-code and MFA-device enrolments on privileged accounts and for tenant-wide Customer Takeout exports, because COLD RELIC's enrolled authenticators survived password and session resets.
Update every TYPO3 instance to 14.3.12, 13.4.40, 12.4.52 ELTS, 11.5.57 ELTS or 10.4.63 ELTS now, and on 14.x configure HTTP/allowed_hosts/dashboard_rss; until then check the public web root and the project root for unexpected files whose name starts with FIRST_INSTALL and the backend for administrator or system maintainer accounts nobody created, because a low-privileged backend user can re-enable the installer and TYPO3 states that one flaw in the same release is actively exploited.
Upgrade every on-premises N-central server to 2026.4 Hotfix 3 (build 2026.4.1.39) and every Agent and Probe to 2026.4.1 HF3 even if already on Hotfix 2, confirm the Legacy API is disabled and not reachable from the internet, and on any server that had the Legacy API on check the web logs for POST requests to the Axis endpoint /dms/services/ServerUI carrying a SOAP fault and the host for a symlink at the N-central flag file path, treating a hit as a full compromise, because public proof-of-concept code exists for all 17 new CVE ids, Hotfix 2 did not stop 4EverAble2Shell and N-able does not say Hotfix 3 fixes it.
Search lockfiles, requirements files and CI install logs for agent-framework 1.26.1 or 1.26.2 installed since 2026-11-17 20:35 UTC, and on every host or job that ran Python after installing either, treat the environment variables, local credential files, Azure managed-identity and Azure DevOps tokens and, for 1.26.2, GitHub Actions OIDC tokens as stolen and rotate them, then pin agent-framework to 1.26.0 or earlier.
Apply Mitel's patch KB000129195 to every MiCollab on 10.3 (10.3.0.18) or earlier, including systems already moved to 10.3.0.18 for the July advisory, and until it is in, restrict external access to the Audio, Web and Video Conferencing component, because Mitel scores one unauthenticated flaw CVSS 10.0 and says the five flaws together amplify the impact.
Block T3 and IIOP to every WebLogic Server from networks that untrusted hosts can reach and apply the November 2026 CSPU first to internet-reachable Access Manager, WebLogic Server, E-Business Suite and APEX or REST Data Services instances, because Oracle marks 56 rows at CVSS 9.0 or above as exploitable without authentication and names blocking the required network protocols as its interim measure for attacks that need no privileges.
Update every Drupal site to 11.4.13, 11.3.22 or 10.6.22 now and move sites on end-of-life lines (Drupal 11.2.x and 10.5.x and below) to a supported branch, and where the update has to wait secure the installer as Drupal's instructions for securing install.php describe, because SA-CORE-2026-017 lets an attacker who reaches the installer attempt a reinstall that can end in site takeover or remote code execution.
If Barracuda notified you that your Managed XDR Cloud or Monitored Endpoint integration credentials may be affected, revoke and recreate each integration credential and review activity in the connected systems (Microsoft 365, Google Workspace, endpoint security platforms and others) since 2026-11-01; Barracuda says an organisation it did not notify needs no action.
Upgrade every Puppet Enterprise server to 2023.8.13, 2025.12.1 or 2026.0.1 (OpenVox Server to 9.2.0 or 8.18.0; open-source Puppet gets no fix) and, until the upgrade is in, stop the CA from autosigning certificate requests without strong verification, because CISA lists the flaw as exploited and any holder of a CA-signed agent certificate reaches code execution on the server.
On every internet-facing Apache HTTP Server that loads mod_http2, update to 2.4.71 or to your distribution's fixed package (Debian bookworm 2.4.71-1~deb12u1, trixie 2.4.71-1~deb13u1 or forky and sid 2.4.71-1, Ubuntu per USN-9090-1, Alpine 2.4.71-r0, RHEL 9 mod_http2 2.0.26-7.el9_9.1, RHEL 10 mod_http2 2.0.29-5.el10_3.1) and restart it; update SUSE Linux Enterprise 15 SP7 to apache2 2.4.71-150700.4.30.1; where no fixed package exists yet, as on the SUSE products still in progress, remove h2 and h2c from the Protocols directive until one does, because Red Hat reports an unauthenticated code-execution path and a public exploit, though it rates its RHEL 9 and 10 packages Important rather than Critical for their harder exploitation.
On every Axis network device on AXIS OS version 6.50.5.10 or higher, set System.HTTP2.AllowH2 and System.HTTP2.AllowH2c to no through param.cgi now (Axis Device Manager mass-deploys it) and install the service release when Axis ships it, then revert the settings, because Axis says an unauthenticated attacker could gain administrator access and has released no fixed AXIS OS.
2026-11-18T0356Z-intel· Sonnet 5.5 · window 26 h · 10 entries published
Verification & coverage notes
Window 26 h (gap 24.0 h to 2026-11-17T0355Z-intel). Ten new entries and three changelog records. The will-publish set was too large for a main-agent deep read, so four scoped cti-research follow-ups (FU1 to FU4, Sonnet 5.5, 04:36:43Z to 05:05:26Z) read the primaries and wrote claim sheets under work/2026-11-18T0356Z-intel/deepread.FU*.md; the entries were composed from those sheets and the saved bodies, and each evidence quote was checked as a literal substring of its saved body.
Backlog: TYPO3 (published, 2026-11-18/typo3-core-security-releases-2026-11-17-installer-chain) and Oracle CSPU (published, 2026-11-18/oracle-november-2026-cspu-56-unauthenticated-critical-rows) were struck as published; the Open table is empty.
KEV: the only in-window addition, CVE-2026-115731 (Perforce Puppet Enterprise), is the entry 2026-11-18/cve-2026-115731-puppet-enterprise-report-deserialization-rce; no covered CVE newly flags ransomware use.
Updates: OptimJack (update: Optimizely confirmed an exploited vulnerability; the entry's 2026-11-13 start for the Optimizely files was wrong and is corrected in place), N-able N-central (update: Hotfix 3 fixes 17 more CVE ids, blocks the Legacy API permanently; the "no fix, follow-up hotfix planned" statements are replaced), Apache httpd CVE-2026-119618 (update: Axis OS 6.50.5.10 and higher affected, no fixed release).
Single-source: COLD RELIC (Mandiant/GTIG first-hand, no independent coverage), Barracuda and RIKEN (victim's own disclosure; their press relays add nothing independent), Drupal PSA, TYPO3, Mitel and Oracle (disclosing vendor), Puppet (CISA carries exploitation alone, no actor or telemetry; the flaw itself is multi-source).
Contradictions surfaced in entries: Java4Shell impact (early source-reading analyses expected a JVM crash, Calif demonstrates code execution, vendors rate CVSS 3.1 8.1 against Calif's 4.0 9.2); TYPO3's "actively exploited" sits on SA-029 (Medium), not on the Critical SA-027 chain, and an automated CISA record says exploitation none; Mitel's range ("10.3.0.18 and earlier") against CERT-FR and BSI ("before 10.3.0.18"); OptimJack's start date for the Optimizely files; Barracuda's ARTEX statement grows in strength across its three pages.
First-pass statements the deep reads corrected: a public trigger proof of concept and a CVE record exist for Java4Shell (S3 said neither); the Drupal PSA never says "proof of concept" (the risk vector carries E:Proof); the Axis PDF never says "HTTP/2 flaw on the web interface"; Oracle's "21 families" and "640 CVEs" are Tenable's counts and only the TopLink row is 10.0 (the T3 and IIOP row is 9.8).
Lookback item: Barracuda (primary 2026-11-10 to 2026-11-13, first coverage from BleepingComputer's 2026-11-16 sighting within its 49 h lookback; event_date 2026-11-10).
Calibration: high on Java4Shell and TYPO3 (OptimJack and the N-able N-central entry stay high); everything else notable. Held at notable after the verifier's iteration-1 challenge: Puppet (exploited per CISA KEV, but constituency use of Puppet Enterprise is not established), Drupal (pre-announcement without CVE, component or fixed version, PSA rates target distribution uncommon and exploit status proof of concept; the update on the release is the point to re-grade) and Apache httpd (Axis finding adds an unauthenticated 9.8 with no fix, but no exploitation is reported and the flaw needs HTTP/2 enabled). No critical: no source shows exploitation of an exposure the constituency demonstrably runs with a time-critical action.
Next fire: the Drupal core security release is due 2026-11-18 17:00 to 21:00 UTC, after this run publishes; the heads-up entry needs the SA-CORE advisory (CVE ids, fixed versions, exploitation status) as an update record. Also re-check ZDI advisories for the Oracle APEX and ORDS rows and Huntress's N-central Hotfix 3 exploitation reports.
borderline-drop: Apache Roller CVE-2026-129448 (S1.2), CNA-stated exploited but a niche blog server with no constituency exposure established.
borderline-drop: ASUS Control Center hard-coded keys (S1.9), niche, single EUVD record, no exploitation.
borderline-drop: Eclipse Mosquitto CVE-2026-110486 (S1.10), websockets listener only, no exploitation, fixed version stated inconsistently (2.1.3 against 2.1.4).
borderline-drop: Brotli vendor-impact long tail (S1.14: Siemens ICSA, Oracle rows, Xerox and Lexmark, Cisco revision 1.14); the covered entry already carries the Siemens and Cisco lists and is over its analysis bound; the Oracle rows are in the CSPU entry.
borderline-drop: Juniper Junos OS Evolved CVE-2026-98432 (S1.15), single CERT-FR relay, vendor page unreadable, fix pending.
borderline-drop: Danfoss AK-SM 800A (S2.3); the covered entry's 2026-11-14 record already carries Midnight Blue, CCB and the BSI position; BSI version 1.1 changes no decision.
borderline-drop: BACS weekly review 46 (S2.4), phishing awareness, no new technique.
borderline-drop: Operation Endgame Gootloader, CastleRAT and StealC (S2.5), takedown with no defender-side decision change.
borderline-drop: FBI and DOJ FaceStealer (S2.6), DPRK crypto-theft social engineering, a class already covered (Binance, Sophos, PolinRider).
borderline-drop: Quantum Systems QBase update server (S2.7), press-reported, no Swiss customer, no vector beyond the update server.
borderline-drop: Reka (S2.8), Swiss private cooperative, no vector or actor, below the incident floor.
borderline-drop: Zurich cloud strategy RRB 1155/2026 (S2.9), cantonal procurement policy with no SOC decision.
borderline-drop: Microsoft 365 outage 12 to 13 November (S2.10), availability, not security.
borderline-drop: Project Zero "Full X" disclosure policy (S2.11), no near-term defender decision.
borderline-drop: Operation ShadowCloud (S3.2), Kaspersky and Sophos corroborate each other, but victims are in Asia and Africa, no source ties it to Europe, and neither post carries the cloud-service detection points the first pass proposed.
borderline-drop: ARTEX delta from Amazon MadPot, Cloudflare and Talos (S3.5), financially motivated campaigns against LLM gateways with no government target, mostly roll-up of covered reporting, and the covered entry already exceeds its analysis bound; the Dosu post-mortem that names ARTEX is a lead for the audit.
borderline-drop: Turla Cereal delta, Romania's SRI statement (S3.6), names Turla only in an opening clause, says only that unspecified Microsoft system vulnerabilities were exploited, and does not say it is the ESET case.
borderline-drop: Fireworks AI two-incident bulletin and Uber (S4.3), no source ties the second incident to the Cursor marketplace compromise; the delta concerns other victims and no Swiss or European public-sector party.
borderline-drop: Microsoft device-code phishing against Copilot CLI (S3.8), laboratory proof of concept, not observed, device-code class already covered.
borderline-drop: Enea SilentAir and LIMINAL PANDA (S3.9), single vendor, Asian operators.
borderline-drop: Infoblox third-party migration breach (S4.4), thin statement, detail rests on anonymous sources relayed by two outlets, no Swiss link.
borderline-drop: Kreiskrankenhaus Greiz-Ronneburg (S4.5), no actor, vector or confirmed attack type, below the incident floor.
borderline-drop: California v. OpenAI (S4.6), legal milestone, no technical delta.
Leads not pursued (below the bar today): N-able Mail Assure and SpamExperts server incident (N-able status page, 2026-11-18 03:05Z, vendor channel, no vector), CNIL's ten fines on French local authorities (2026-11-09, outside the lookback), Maileva (freshest source 2026-11-16, no Swiss client).
Included with reduced confidence: none.
Coverage gaps: belden-psirt (403), connectwise-trust-center (403), sygnia (403) and lab52 (no route) were not in any slice; perforce-portal (script-rendered page), juniper-psirt (script-rendered page), europol-newsroom (React shell, head metadata only), drupal.org page URLs (client challenge to extract, read through the PSA feed and WebFetch), justice.gov (Akamai interstitial), the Maine and Oregon breach filings behind the Uber count (not read, count not carried), mainichi.jp (paywall).
Operational: S3 reported two scratch files written at the filesystem root through a mis-scoped shell variable and moved to its scratchpad; nothing in the repository was touched outside work/. discovered_at stamps were set to the real composition clock. Mid-run stop-hook nudges to commit were not acted on before Phase 6.
Verification: five iterations (NEEDS_FIXES at 1 to 5: 32, 10, 7, 4 and 2 findings). Iteration 5 returned two low-confidence F3 findings and no F1 or F4, so the loop ended by the early exit (truth + editorial = 2); both findings were fixed after the pass and not re-verified by a further iteration, hence verification_residual_count: 2. The three F16 priority challenges from iteration 1 were declined and not re-raised. No CLEAN verdict was reached, so there is no double-CLEAN; the entries ship on the early exit.
Source health: the sweep left one UNSOLVED row, typo3-security-advisories (candidate added this run), verdict irrelevant: the feed works and is current (newest item 2026-11-17), but its titles carry only two of the health check's relevance terms (RELEVANCE_MIN is 3) because the vocabulary lacks words such as advisory and forgery. Left for the quality audit to widen the vocabulary or accept the verdict; no tool was changed in this fire.