Koffei

Daily CyberNews · 06/11/2026

‹ 05/11/202609/11/2026 ›
Today in 30 seconds
  • 🔴 CVE-2026-102255 in SonicWall SMA 1000 (CVSS 10.0) is actively exploited days after disclosure.
  • 🔴 Upgrade Rails to 8.1.4.1 or 8.0.5.2: the Active Storage file-write flaw has a real attack report.
  • 🔴 Take refrigeration controllers off the internet: CISA and FBI warn as European attacks continue.
  • 🔴 Apply Apple's 2024 fix for CVE-2024-27834: in KEV, used by APT29, deadline Nov 18.
  • 🟠 Update OpenSSH to 10.6p2 on Alpine systems running sshd without PAM.

Vulnerabilities 9

🔴 Exploited or active

CISA adds old Apple PAC-bypass bug to KEV over APT29 abuse

CVE-2024-27834AppleCVSS 5.5in KEVAPT29

WhatCISA added the 2024 Apple Pointer Authentication bypass to KEV, citing APT29 use in a zero-click chain delivering info-stealing malware.

DoApply Apple's May 2024 updates; federal agencies must patch by Nov 18, 2026.

Details and sources
CISA added CVE-2024-27834 (improper authentication allowing a Pointer Authentication bypass in Apple's Safari, iOS/iPadOS, macOS, tvOS and watchOS, CVSS 5.5, patched by Apple back in May 2024) to its Known Exploited Vulnerabilities catalog on Nov 4, 2026, citing active exploitation attributed with "moderate to high confidence" to APT29 (Cozy Bear/Midnight Blizzard) as part of a zero-click chain delivering info-stealing malware; federal agencies must patch by Nov 18, 2026. The Hacker News (source)
🟠 High impact

Zyxel's own patch doesn't close its GS1900 switch zero-day

CVE-2026-107441Zyxel GS1900patch fails

WhatAn out-of-bounds write lets attackers run root commands on ten GS1900 models, and a working exploit still beats Zyxel's Oct 23 Patch 3.

Details and sources
CVE-2026-107441 is an out-of-bounds write in the web-management handler of ten Zyxel GS1900 switch models (firmware 2.90 and earlier) that lets an attacker run commands as root; Binarly found a working exploit existed since Sept 29, 2026 and that it still works against Zyxel's own Oct 23 "Patch 3," meaning GS1900 switches remain exploitable on the latest available firmware. A related, already-patched GS1900 bug (CVE-2026-7273) was exploited against nearly 1,000 switches in August and added to KEV in September. The Hacker News (source) · SecurityWeek (source)
🟠 High impact

Critical MITM flaw in Debian/Ubuntu's APT package manager

CVE-2026-109518APTDebian/UbuntuCVSS 7.7

WhatA network attacker controlling HTTP traffic can abuse APT's retry logic to swap a verified package for code that runs as root.

DoInstall the fixed APT builds shipped Nov 4 for Debian and all 8 tracked Ubuntu releases.

Details and sources
CVE-2026-109518 (CVSS 7.7) lets a network attacker who can manipulate HTTP traffic — e.g. via a malicious mirror or proxy — abuse APT's download-retry logic to swap a verified package for attacker-controlled code that runs as root; exposure is on by default from Debian 12 / Ubuntu 22.04 onward. No exploitation reported yet. Fixed builds shipped for all 8 tracked Ubuntu releases and for Debian on Nov 4, 2026. The Hacker News (source)
🔴 Exploited or active

SonicWall: one flaw already under attack, two more it won't fix

CVE-2026-102255SonicWall SMA 1000CVSS 10.0exploited

WhatA pre-auth SSRF in SMA 1000 gateways is actively exploited, and SonicWall will not fix CVE-2026-109593 or CVE-2026-109594 in end-of-life GMS.

DoPlan to retire end-of-life SonicWall GMS, since it will get no security update for these flaws.

Details and sources
CVE-2026-102255, a pre-auth SSRF in SonicWall's still-supported SMA 1000 gateways (CVSS 10.0), is being actively exploited just days after disclosure. Separately, SonicWall disclosed CVE-2026-109593 (unauthenticated file read, CVSS 9.1) and CVE-2026-109594 (post-auth SQL injection, CVSS 6.8) in its Global Management System — but says GMS is end-of-life and "no remediation or security update will be released," leaving any organization still running it permanently exposed. The Hacker News (source) · BleepingComputer (source)
🔴 Exploited or active

Critical file-write flaw in Rails Active Storage

Rails Active StorageGHSA-f3w6-w445-fwh9CVSS 9.8

WhatA crafted ImageMagick upload can let attackers write arbitrary files on the server through Active Storage; Rails has one real-world attack report.

DoUpgrade Rails to 8.1.4.1 or 8.0.5.2.

Details and sources
a crafted ImageMagick upload can let an attacker write (and potentially read) arbitrary files on the server through Rails' Active Storage component (GHSA-f3w6-w445-fwh9, CVSS 9.8; no CVE assigned yet). Rails says it has received one real-world attack report and doesn't assume it's the only one. Fixed in Rails 8.1.4.1 and 8.0.5.2, released Nov 4, 2026. The Hacker News (source)
🟠 High impact

OpenSSH emergency fix for Alpine Linux auth bypass

OpenSSHAlpine Linuxmuslauth bypass

WhatA password check quirk with musl libc lets long passwords bypass SSH authentication on Alpine systems running sshd without PAM.

DoUpdate to OpenSSH 10.6p2 or Alpine's backported fixes.

Details and sources
a password-check quirk interacting with musl libc let long passwords bypass SSH authentication on Alpine-based systems running sshd without PAM; OpenSSH calls it "critical" but has not assigned a CVSS or CVE yet. Fixed in OpenSSH 10.6p2, with Alpine backporting fixes to all supported branches back to 3.16. The Hacker News (source)
🟡 Worth knowing

Microsoft Paint RCE fix exists — but admins can't actually get it

Added after publishingCVE-2026-83844Microsoft PaintCVSS 8.1

WhatAn out-of-bounds read in Paint could allow code execution via a crafted image, but the Store kept serving the vulnerable 11.2605.81.0 build.

DoCheck the installed Paint version with Get-AppxPackage Microsoft.Paint and confirm it is not 11.2605.81.0.

Details and sources
CVE-2026-83844 is an out-of-bounds read (CWE-125) in Microsoft Paint, CVSS 8.1, that could let an attacker achieve remote code execution via a specially crafted image file; no network privileges are needed, though attack complexity is high. Microsoft's advisory originally listed Paint version 11.2605.91.0 as the fix, but IT admins report the Microsoft Store keeps serving the older, still-vulnerable 11.2605.81.0 build, and Microsoft's Security Update Guide has since switched the listed fix version to 11.2609.401.0 — adding to the confusion over whether systems are actually patched. No in-the-wild exploitation is confirmed; a public GitHub script checks installed Paint versions against the advisory but is a detector, not a working exploit. Admins can check their installed version with Get-AppxPackage Microsoft.Paint | Select-Object Name, Version. Windows Report (source)
🟡 Worth knowing

Update — Paint fix is rolling out, but the CVE record still doesn't match

Added after publishingCVE-2026-83844Paint 11.2609.401.0Microsoft Store

WhatThe Paint fix is build 11.2609.401.0, delivered only via the Microsoft Store, while the CVE and NVD records still list mismatched versions.

DoMake sure Paint is 11.2609.401.0 or later, push it with Intune, and check per-user and provisioned copies.

Details and sources
the fix for CVE-2026-83844 is build 11.2609.401.0, delivered only through the Microsoft Store (not Windows Update); checks on Nov 6 found it as the newest build in all ten countries tested, and the admin who first reported being stuck on 11.2605.81.0 has since received it via Store auto-update, though rollout speed to other machines is unclear. The CVE record and NVD entry still list only versions below 11.2605.91.0 as affected — a discrepancy with the advisory's actual fix build that hasn't been corrected — so Microsoft's practical guidance is that any copy at 11.2609.401.0 or later is safe under either version number. Fleet admins can push it via Intune's Store app update scan and should check both per-user and provisioned Paint copies. The Hacker News (source)
🟡 Worth knowing

PoC: a malicious browser extension could hijack Claude's Dispatch feature to run code on your computer

Added after publishingClaude Dispatchbrowser extensionPoC

WhatResearchers showed a browser extension with script access to claude.ai could abuse Dispatch to run code on a user's computer; no real attacks seen.

Details and sources
researchers at Forever Security (writeup by Gal Weizman) disclosed — as a proof-of-concept, with no real-world attacks observed — that a browser extension with script access to claude.ai could abuse Dispatch, Anthropic's feature for sending tasks from a phone to Claude Desktop, to send instructions that looked like they came from the signed-in user and auto-click through Claude's own approval prompts every second. In testing this got Claude to install an npm package whose install script opened the Calculator app, with no browser exploit needed; the researchers say the same access could steal files or install a backdoor, including reaching a second, cross-device-linked Claude Desktop instance. Anthropic accepted the report as valid in May but closed it as a duplicate of an internal finding, and reportedly told the researchers Dispatch would be retired by end of October — though its help page still describes it as available "for now" to some existing Pro/Max/Ultra users. The researchers warn the same design issue could extend to Claude Code Remote Control and Claude Cowork sessions, which they plan to test next. The Hacker News (source)

Breaches / Incidents 5

🟠 High impact

Chainalysis confirms its agentic-AI breach exposed data for every user

Added after publishingChainalysisbreachall users

WhatChainalysis found its agentic-AI intrusion exposed names, usernames, password hashes and live tokens for all users.

DoChainalysis users must reset passwords and rotate API keys.

Details and sources
following up on an intrusion it first disclosed Oct 30, blockchain-analysis firm Chainalysis says its investigation found the attacker — who used offensive agentic AI techniques to gain unauthorized access — exposed names, usernames, password hashes and live tokens for all users, not just the customer-password subset first reported. Chainalysis says it has no evidence investigation data was accessed or that its attribution data was tampered with; it pulled Reactor, Data Solutions, the KYT interface/API and other screening products offline during the response and is requiring a full password reset and API-key rotation. No attacker has been named; the company likens the intrusion to a recent wave of agentic-AI attacks on other tech platforms (Chainstack, OpenCode, Warp, WorkOS). The Hacker News (source)
🟠 High impact

French bank Crédit Mutuel Arkéa hit for an estimated $133M+

Crédit Mutuel ArkéaFrance$133M+

WhatAttackers compromised a third-party provider's platform and used the bank's own account to send fraudulent transfers, estimated at over $133 million.

Details and sources
attackers compromised a third-party service provider's platform to reach the bank directly, then used the bank's own account to send fraudulent transfers, apparently to launder funds into crypto. Blockchain investigator ZachXBT estimates more than $133 million was stolen; the bank won't confirm the figure but says it can "validate the order of magnitude." Customer accounts and assets were not affected, and the bank says the loss doesn't threaten its solvency. No attacker group has been named. The Record (source)
🟡 Worth knowing

Suspected Russian attack hijacks Ukrainian outlet Kyiv Independent to spread fake Zelensky nuclear-strike claim

Kyiv IndependentUkrainesuspected Russian actors

WhatHackers took over the outlet's CMS and newsletter to spread false messages, including a fake claim of a Zelensky nuclear strike.

Details and sources
hackers took over the outlet's CMS and newsletter system to post and email false messages, including a fabricated claim that President Zelensky launched a nuclear strike on Kyiv. No banking or donor payment data was touched (handled by a separate processor). The outlet attributes the attack to "suspected Russian actors" based on message content, though no technical attribution evidence has been published. The Record (source)
🔴 Exploited or active

Supermarket refrigeration attacks spread to a second wave across Europe

DanfossCopelandDixellEurope

WhatA second wave of cooling-system attacks hit retailers in nine European countries overnight into Nov 2, and CERT-FR says the campaign is probably ongoing.

Details and sources
overnight into Nov 2, 2026, cooling-system attacks hit retailers in Belgium (Delhaize, Carrefour, Intermarché, Spar, an IKEA store), Sweden (ICA, Coop), Ireland (Musgrave/SuperValu), Estonia (Selver), the Netherlands, Finland (Kesko), Portugal (Intermarché) and Italy (Coop/Radenza), mostly via Danfoss control systems — the same vendor implicated in October's ~700-store French attack. No group has claimed responsibility; Belgium's Centre for Cybersecurity says attribution "may never be determined," and its federal prosecutor has opened an investigation. Impact per chain stayed small (a few dozen stores at most), with some food losses, but CERT-FR warns the underlying campaign against Danfoss, Copeland and Dixell controllers is probably still ongoing. The Hacker News (source) · Help Net Security (source)
🔴 Exploited or active

CISA and FBI warn on industrial refrigeration security — and a French blood bank's cold storage was also hit

Added after publishingCISAFBIindustrial refrigerationFrench blood bank

WhatUS agencies warned operators of refrigeration and cold-chain systems after the European attacks, and France's blood service said three cold-storage sites were compromised.

DoTake refrigeration and cold-chain monitoring systems off the internet, change default passwords, patch firmware and replace unsupported devices.

Details and sources
on Nov 5, 2026, CISA, the FBI and three other federal agencies urged operators of industrial refrigeration and cold-chain monitoring systems to take them off the public internet, change default passwords, patch firmware and replace unsupported devices, citing risk from authentication-bypass and code-injection flaws or weak configurations that could force equipment into continuous heating/defrost cycles. The advisory was prompted by the European supermarket refrigeration attacks above; separately, France's national blood service confirmed that monitoring/control systems at three of its cold-storage sites were also compromised, though staff intervened before any blood products or samples were affected. No link between the blood-storage incident and the same attackers has been confirmed. Malwarebytes (source)

Malware / Ransomware 5

🟠 High impact

PAYLOAD and HYFLOCK — two new ransomware operations surface

PAYLOADHYFLOCKnew ransomware

WhatPAYLOAD uses Group Policy for encryptionless extortion and also has an ESXi binary, while HYFLOCK is a new RaaS whose buggy encryption can destroy files.

Details and sources
PAYLOAD is a Group Policy Object-based "encryptionless extortion" attack that hit a Middle East manufacturer via domain admin-equivalent access (disabling local admin accounts, forcing ransom-note logon banners), alongside a separate PAYLOAD binary built to target VMware ESXi; HYFLOCK is a new ransomware-as-a-service operation with 13 listed victims whose buggy encryption can permanently destroy parts of larger files. The Hacker News (source)
🟠 High impact

WKLP infostealer/RAT targets US and French government networks

WKLPinfostealer RATUS and French government

WhatA Python infostealer and RAT disguised as MoUsoCoreWorker.exe steals credentials and files, adds Defender exclusions and kills security processes.

DoHunt for a fake MoUsoCoreWorker.exe, unexpected Defender exclusions and disguised firewall rules.

Details and sources
a Python-based infostealer and remote access trojan disguised as the Windows process "MoUsoCoreWorker.exe" steals credentials, browser data and files, adds Microsoft Defender exclusions, kills security processes and hides behind disguised firewall rules to persist. The Hacker News (source)
🔴 Exploited or active

Backdoored copies of the ARTEX AI pentesting tool caught abusing CAPTCHA field for root access

ARTEXbackdoored Docker imageCrafty Falcon

WhatA fork of the ARTEX AI pentesting tool pulls a backdoored image granting a root shell via a secret CAPTCHA marker; 62 consoles served it.

DoHunt for ARTEX consoles running the backdoored Docker build from the altered fork.

Details and sources
after abuse of the open-source "ARTEX" AI pentesting tool led its own developer to pull its GitHub repo, a fork altered its Docker build to pull a backdoored image (~9,000 downloads) that grants a root shell when a secret marker is typed into the CAPTCHA field, bypassing credential checks entirely. 62 internet-facing consoles were serving the backdoored build Nov 2–4, 2026. India's CERT-In separately named an AI-assisted actor, "Crafty Falcon," using ARTEX and similar autonomous tools against Indian banks and crypto exchanges. The Hacker News (source)
🟡 Worth knowing

Trojanized Codex jailbreak bundle injects ads into AI responses

Check PointOpenAI Codexprompt adware

WhatA Chinese Codex GUI tool bundles a jailbreak that installs hacking skills and injects ads for a third-party LLM gateway into every response.

Details and sources
Check Point Research found "prompt adware" bundled inside a popular Chinese GUI tool for OpenAI's Codex, distributed via a GitHub repo with 403 stars; it jailbreaks the model via role-play framing, installs pentesting/password-cracking "skills," and persistently injects ads for a third-party LLM gateway into every response — even a simple "hi." The Hacker News (source)
🟠 High impact

Three branded kits abuse Microsoft's device-code sign-in flow

SkyW4lkerGeorgeGinxMir0Authdevice code phishing

WhatThree phishing kits abuse Microsoft 365's device-code sign-in flow to steal account access without passwords, targeting US government, banking, healthcare and other sectors.

Details and sources
SkyW4lker, GeorgeGinx and Mir0Auth are distinct phishing-as-a-service kits that all abuse Microsoft 365's device-code authentication flow to steal account access without ever touching a password — routing victims through fake contracts, hijacked Adobe Shared Links, or fake DocuSign pages. Targets span US government, banking, healthcare, education and tech sectors. The Hacker News (source)

Threat Actor Activity 3

🔴 Exploited or active

APT29 hijacks a Ukrainian web-analytics script to deliver iOS/Android N-day exploit chains

APT29PlerdyCVE-2024-27834

WhatAPT29 used compromised sites, lookalike domains and a hijacked Plerdy analytics script to serve iOS and Android exploit kits that steal cookies and messages.

Details and sources
Sekoia reports a campaign "presumed" to be run by Russia-linked APT29 (Cozy Bear), combining compromised sites, lookalike domains and a supply-chain compromise of Ukrainian analytics service Plerdy SEO Tracker (malicious code live Aug 27–Sep 3, 2026) to serve an iOS exploit kit ("DarkSword") and an Android one ("AndroSnitch") that steal cookies, messages, Wi-Fi passwords and more. Targets include a US chemical-industry trade association, a European far-right party, and a Bulgarian military-related site, with lookalikes aimed at Ukraine, Czechia and Bulgaria. The iOS chain's use of the now-KEV-listed CVE-2024-27834 prompted CISA's Nov 4 catalog addition (see Vulnerabilities). The Hacker News (source)
🔴 Exploited or active

India's CERT-In flags "Crafty Falcon," an AI-assisted attacker targeting banks and crypto exchanges

Crafty FalconCERT-In CIAD-2026-0058BFSI

WhatIndia's CERT-In flagged an AI-assisted actor using autonomous tools including backdoored ARTEX builds against Indian banks and crypto exchanges.

Details and sources
the actor reportedly uses autonomous offensive-AI tools — including the backdoored ARTEX builds noted above — to run full attack chains "at machine speed" and rotate infrastructure within hours, targeting Indian BFSI firms and crypto/digital-asset exchanges; CERT-In issued advisory CIAD-2026-0058 and new AI Security Auditing Guidelines. No arrests have been reported. The Hacker News (source)
🟡 Worth knowing

Canadian national extradited over years-old US school hacking/cryptomining scheme

Ryan James Roachextraditioncryptomining

WhatA Canadian was extradited to the US over a 2016 to 2019 scheme hacking US schools to mine cryptocurrency; a second suspect awaits extradition.

Details and sources
Ryan James Roach was extradited from Canada and arraigned in Rochester, NY, over a 2016–2019 scheme that hacked into US educational institutions (including machines with supercomputers) in New York, Tennessee, North Carolina and Massachusetts to mine cryptocurrency; a second Canadian, Mathiew James Stubbings, is still awaiting extradition. Both face up to 10 years if convicted. The Record (source)
‹ 05/11/202609/11/2026 ›