Koffei

Daily CyberNews · 16/11/2026

‹ 13/11/202617/11/2026 ›
Today in 30 seconds
  • 🔴 Patch Check Point Gaia today: CVE-2026-123039 is exploited after PHP2Shell and the KEV deadline is now.
  • 🔴 Update UnrealIRCd to 6.2.7.1: CVE-2026-114713 is exploited and its KEV deadline is today.
  • 🔴 Second SCKit wave hit 205 ScreenConnect hosts: disable the flagged extensions and run clients on 26.6.5 or later.
  • 🔴 FBI warns ARTEX AI agents power attacks across sectors: rate-limit sensitive APIs and watch for request bursts.
  • 🟠 Take Danfoss AK-SM800A controllers off the internet: six unpatched flaws allow remote root without login.

Vulnerabilities 3

🔴 Exploited or active

Check Point Gaia root flaw: CISA deadline is today, and a stopgap exists

CVE-2026-123039CVSS 7.8in KEVCheck Point Gaia

WhatA local root flaw in Check Point's DynamicCLI daemon was exploited against about two dozen customers after PHP2Shell initial access; the KEV deadline is today.

DoInstall JHF R81.20 Take 174, R82 Take 132, R82.10 Take 52 or R82.20 Take 14, or stop dynamic_server with dbset until patched.

Earlier coverage: 13/11
Follow-ups: 17/1119/11
Details and sources
CVE-2026-123039 (CVSS 3.1 7.8, local, CWE-732/CWE-862, EPSS about 2%) lets any local user send commands to the root DynamicCLI daemon through the world-writable /tmp/dynamic_socket. It hits Quantum Security Gateways, Management and Log Servers, Multi-Domain and Maestro on R81.20 up to Jumbo Hotfix Take 173, R82 up to Take 130, R82.10 up to Take 50 and R82.20 up to Take 10; fixes are JHF R81.20 Take 174, R82 Take 132, R82.10 Take 52 and R82.20 Take 14, while R81 and R81.10 are end of support and get no fix (advisory sk1000362). Check Point saw it exploited against "a couple of dozen customers" after initial access through the PHP2Shell flaws CVE-2026-109618 and CVE-2026-109619 (CVSS 9.8, in KEV since 26 October). CISA added it to KEV on 13 November with a 16 November deadline. If you cannot patch yet, dbset process:dynamic_server then dbset :save stops the service (and the Dynamic CLI commands); check /var/log/dynamicCli.log for commands without the Clish wrapper and confirm the socket is 0660 after patching. Spark firewalls and Smart-1 Cloud are not affected. Cyber Security News (source) · Security Affairs (source) · OpenCVE (source)
🔴 Exploited or active

UnrealIRCd label-injection flaw: exploited, KEV deadline today

CVE-2026-114713CVSS 8.6in KEVUnrealIRCd

WhatAn UnrealIRCd label-injection flaw lets a user on linked networks inject server commands such as KILL or GLINE, and attackers are exploiting it.

DoUpdate UnrealIRCd to 6.2.7.1 or apply hot-patch bundle-20261107; federal agencies must patch by today.

Earlier coverage: 13/11
Details and sources
CVE-2026-114713 (CVSS 3.1 8.6, network, no authentication, CWE-140) in the labeled-response module does not escape the IRCv3 "label" tag, so a user on a network of two or more linked servers can inject server-to-server commands such as KILL or GLINE. Versions 6.0.0 before 6.2.7.1 are affected (the end-of-life 5.x line too); fix with 6.2.7.1 or hot-patch bundle-20261107. Attacks began about a day after the 7 November hotfix, and federal agencies must patch by today. The Hacker News (source) · OpenCVE (source)
🟠 High impact

Six unpatched Danfoss AK-SM800A flaws allow remote root without login

Added after publishingDanfoss AK-SM800Ano CVEno patchICS

WhatSix unpatched flaws in the Danfoss AK-SM800A refrigeration controller allow remote root code execution without login; no exploitation is known.

DoRemove internet exposure including 4G/5G modems, use a VPN for remote access, and check setpoints, defrost schedules, alarms and accounts.

Earlier coverage: 10/1109/1106/11
Details and sources
Carlo Meijer of Midnight Blue found six flaws in the Danfoss AK-SM800A store refrigeration controller in under an hour of manual review: five in web request handlers that need no login and run as root, and a sixth in the Strict mode gateway, where a crash that could lead to code execution was shown (the gateway lacks stack canaries). An attacker could change setpoints, trigger defrost cycles, hide activity from logs or plant persistent code (there is no secure boot). Every version up to 5.3.2 is affected (root code execution shown on 3.3.0); 5.4.0 is disputed: the researchers say it fixes nothing, and Germany's BSI says it cannot rule out that it is affected. No CVE IDs, no CVSS scores, no patch and no Danfoss advisory yet; BSI has no information on exploitation, and nothing links these flaws to the recent supermarket refrigeration attacks. Belgian, Irish and German agencies advise removing internet exposure (4G/5G modems included), using a VPN for remote access and checking setpoints, defrost schedules, alarms, accounts and remote-access settings; firmware updates, password changes and Strict mode are not enough on their own. The Hacker News (source)

Breaches / Incidents 6

🟠 High impact

Cursor confirms a stolen Fireworks AI key led to the poisoned Claude Code extension; Uber and Notion also hit

CursorFireworks AIClaude Code extensionUberNotion

WhatA stolen Fireworks AI key let attackers copy Cursor data and plant backdoored Claude Code extension builds in Cursor's marketplace cache for about 11 hours.

DoCheck whether Claude Code extension builds 2.1.312 or 2.1.313 came from Cursor's marketplace, and look for an unexpected persistent SSH backdoor.

Earlier coverage: 11/1102/11
Details and sources
Cursor says attackers used an access key held by its AI provider Fireworks AI to copy a read-only bucket the two shared, mostly 2025 coding-agent traces from users without Privacy Mode, some with unredacted credentials. With that access they planted Claude Code extension builds 2.1.312 and 2.1.313 in Cursor's marketplace cache on 29 October for about 11 hours; the builds tried to install a persistent SSH backdoor, and up to 5,000 users may have received them. Fewer than 0.05% of users had traces exposed and fewer than 0.01% of customers had API credentials compromised; archived non-production source code, support data and usage metrics were also taken. Anthropic's own extension and Open VSX were not compromised. Cursor has moved inference off Fireworks, rotated credentials and rebuilt its marketplace. Uber is notifying drivers that licence data was stolen through the same Fireworks breach (2,567 in Massachusetts, 3,430 in South Carolina, 1,873 in Rhode Island so far), and Notion says AI chat transcripts from a few workspaces were exposed. No group has claimed it. BleepingComputer (source) · The Hacker News (source) · TechCrunch (source)
🟠 High impact

New extortion group "n0n" claims nearly a billion Expedia traveler records

Expedian0nQubolesupply chain

WhatThe extortion group n0n claims nearly a billion Expedia traveler records; Expedia confirmed access through a compromised third-party data platform but not the figures.

Earlier coverage: 10/11
Details and sources
the Russian-speaking group, which steals data without ransomware, claims 830 million loyalty records from 35 programs, 58 million traveler profiles, 333 million account and credential records, about 2.5 million card-spending records and source code. Expedia confirmed attackers reached its systems through a compromised third-party data platform in a supply-chain attack that hit other companies too, but has not confirmed the figures; Wiz and Sentra tie the attack to a stolen vendor key from Qubole. The scale is the attackers' claim only. Security Affairs (source)
🟠 High impact

La Poste's Maileva breach exposes 22 million recipients in France

La PosteMailevaFrance22 million

WhatA September attack on La Poste's Maileva platform exposed names and postal addresses of about 22 million recipients in France.

Details and sources
a mid-September attack on Maileva, the Docaposte mail and document platform, exposed the names and postal addresses of about 22 million people who received payslips, registered letters or invoices through it (bank details not affected, the company says). Affected senders include social housing providers, Inovie group laboratories and Foch Hospital patients via Kiwee.care, whose payment reminders showed invoice numbers, treatment dates and amounts. Maileva notified the CNIL and filed a criminal complaint; a judicial investigation is under way. Security Affairs (source)
🟠 High impact

Osaka: over a million email addresses leaked through mail provider Arara

OsakaArara77 million addresses

WhatAttackers reached mail provider Arara, exposing over a million Osaka addresses and 77 million across nearly 400 customers, then sent phishing from it.

Details and sources
the city says hackers reached Arara's email distribution service, used by its contractor TKC, exposing addresses of anyone emailed by its online services platform between July and 31 October. Arara says the 2 November attack exposed 77 million addresses across nearly 400 customers, Sega among them, and was used to send phishing mails posing as card companies and banks. The Record (source)
⚪ Context

Japan's emergency cybersecurity package after a wave of breaches

Japanemergency packageactive cyber defense

WhatJapan announced a 10 billion yen cybersecurity package, including wider active cyber defense powers and joint support teams, after a wave of breaches.

Details and sources
Prime Minister Sanae Takaichi's government announced on 6 November a 10 billion yen (about $63 million) package after dozens of companies and agencies suffered breaches: wider "active cyber defense" powers to reach attackers' servers (legal since October), a joint Defense Ministry, police and National Cybersecurity Office support team, a METI "top gun" team using AI by the end of November, and public-private "cyber rescue" teams for hit companies. Takaichi cited the speed of AI-generated attacks. AP (source)
🟡 Worth knowing

LEGO's BrickLink still offline after "unusual activity"

LEGOBrickLinkoffline

WhatLEGO's BrickLink marketplace has been offline since about Nov 5 after unusual activity; no hack or data exposure is confirmed.

Details and sources
the marketplace has been down since about 5 November; the planned 11 November return date was removed, and no hack or data exposure has been confirmed. Malwarebytes (source)

Malware / Ransomware 4

🔴 Exploited or active

SCKit: second wave of rogue ScreenConnect clients, and no proven link to the disabled extensions

SCKitScreenConnectCVE-2026-84869in KEV

WhatA second wave of rogue ScreenConnect clients hit 205 hosts at 41 organisations on Nov 13, and the link to the disabled extensions is unproven.

DoDisable the five flagged extensions, run technician clients on 26.6.5 or later, and hunt for 185.241.211[.]197 and ns.mo7sen[.]dev.

Earlier coverage: 13/1112/11
Follow-ups: 18/11
Details and sources
Huntress saw a new spike from 16:12 UTC on 13 November hitting 205 hosts at 41 organisations, with the rogue client hosted at 185.241.211[.]197 (bulletproof host 1337 Services), relay 007[.]from-ak[.]com:88 and a new c=ExploitedNOTUSA installer value; no follow-on execution of the decoder was seen. The first wave on 11 November hit 288 endpoints at 48 organisations in about 90 minutes. SCKit runs from ScreenConnect's Run Command (sc.msi, sckit-run.bat, a PowerShell decoder from ns.mo7sen[.]dev), installs two rogue clients and uses its RelayTrap module to steal technicians' session cookies and push SYSTEM commands to other managed endpoints. Blackpoint Cyber found no technical link between SCKit and the five third-party extensions ConnectWise disabled (Huntress ties the activity to RMM+ 1.0.31/1.0.32, which accepted a blank access key). ConnectWise asks partners to disable those extensions and run technician clients on 26.6.5 or later, which fixes CVE-2026-84869 (CVSS 9.9, client file transfer and execution without host confirmation, in KEV since 11 September). Huntress (source) · The Hacker News (source)
🟠 High impact

Fake Zillow review pages push NetSupport RAT through ClickFix

NetSupport RATClickFixfake Zillow pages

WhatFake Zillow apartment review pages on .sbs domains trick visitors into running a PowerShell command that installs NetSupport RAT.

DoHunt for traffic to 89.34.90.203, asgomd[.]com and asgrsv[.]com and for NetSupport RAT in AppData with a Startup shortcut.

Details and sources
EfficientIP found "apartments-review" pages on .sbs domains showing a fake verification check that copies a PowerShell command to the clipboard and tells the visitor to run it. The command pulls a script from 89.34.90.203 (written as the decimal 1495423691), downloads a Python runtime, unpacks NetSupport RAT into AppData and adds a Startup shortcut; gateways asgomd[.]com and asgrsv[.]com. The domains date from August; delivery is probably rental-scam links. No attribution. Help Net Security (source)
🟠 High impact

SANSAR: Rust cryptominer hides on AI GPU servers behind a rootkit and fake GPU readings

Added after publishingSANSARXMRigRayGPU servers

WhatREF3364 entered AI GPU servers through an exposed Ray job agent and deployed SANSAR miners hidden by a kernel rootkit and fake GPU readings.

DoCheck for exposed Ray job agents and hunt for the acpi_thermal_mgr.ko rootkit and outbound TCP 8080 backdoor connections.

Details and sources
Elastic Security Labs tracks REF3364, a financially motivated cluster seen in late October that got in through an exposed Ray job agent, used Ray jobs to inventory GPUs with nvidia-smi, then ran a Python reverse shell and added an SSH key. SANSAR, a Rust orchestrator with a Tor onion C2, manages an XMRig CPU miner ("kurt"), an SRBMiner-MULTI GPU miner ("tilki"), the lyrebird Tor transport, an ftrace kernel rootkit (acpi_thermal_mgr.ko, reloaded at boot through DKMS and initramfs) that hides processes, files and sockets, and "netns", a systemd backdoor that reads its controller address from an Ethereum smart contract (0x43eb291f56688c63c93e8a93c4591fc997cf6349) and connects on TCP 8080. A modified NVML wrapper reports 0% GPU use with fake temperature (36-46°C) and power (100-200 W) figures. No CVE is involved; attribution is unclear (Turkish-language comments in the rootkit source). The Hacker News (source)
🟠 High impact

Retadup worm resurfaces seven years after its takedown, now pushing CoffeeLoader

Added after publishingRetadupCoffeeLoaderSmokeLoader

WhatRetadup worm command servers were re-registered and now push CoffeeLoader, blocked by Kaspersky on about 15,000 systems.

Details and sources
Kaspersky says command servers of Retadup, the USB-spreading Windows worm French police disrupted in August 2019, were re-registered about two years ago and first served a SmokeLoader downloader, then from early 2025 CoffeeLoader. The new variant uses an NSIS installer that drops a patched 7-Zip DLL run through rundll32, with shellcode from a new Armoury packer version that now decrypts on the GPU through OpenGL; it persists through a Run key (with a registry block on disabling autorun in Task Manager) and injects into dllhost, WerFault, svchost, OpenWith or AtBroker. Kaspersky blocked it on about 15,000 systems, mostly in Peru, Angola, Russia and Bolivia; who revived the servers is unknown. Retadup has delivered miners, STOP ransomware and the Arkei stealer in the past. The Hacker News (source)

Threat Actor Activity 3

🟠 High impact

BO Team used Log4Shell on vCenter to own a Russian software maker for months

BO TeamCVE-2021-44228in KEVvCenter

WhatBO Team used Log4Shell on a VMware vCenter server to compromise a Russian software maker for months, reaching domain admin and the code-signing server.

DoHunt for an hourly pgcache cron script and connections to nofrillsdns[.]com domains.

Earlier coverage: 13/11
Details and sources
Kaspersky ICS CERT says BO Team (Black Owl, Hoody Hyena, Lifting Zmiy), a group it sees as Ukrainian-origin and focused on espionage and hacktivism, got in through an exposed SSH service, then used CVE-2021-44228 (Log4Shell, CVSS 10.0, in KEV since 2021) on a VMware vCenter 7.0.3 server for a root shell. An hourly "pgcache" cron script posing as PostgreSQL pulled a GSocket backdoor from nofrillsdns[.]com domains, switchable through a DNS record. The attackers dumped LSASS and ntds.dit, took domain admin, reached GitLab and the code-signing server (no malicious release was confirmed) and later tried phishing that passed SPF and DKIM. The Hacker News (source)
🔴 Exploited or active

FBI FLASH: ARTEX AI pentest framework behind attacks in many sectors

ARTEXFBI FLASHAI pentest framework

WhatThe FBI says several actors use the ARTEX autonomous pentest framework in attacks across sectors, some reaching privileged access in under three hours.

DoApply rate limits on sensitive APIs, watch for bursts of automated requests, and check the IOCs in the FBI alert.

Earlier coverage: 13/1112/1111/11
Details and sources
the FBI says several threat actors have used ARTEX, a Go-based autonomous pentest framework that drives external LLMs through recon, exploitation, privilege escalation and data theft, since at least August. More than 1,600 consoles were exposed online in early November, some intrusions reached privileged access in under three hours, and targets include finance, government, technology, defense and healthcare. The original GitHub project was removed on 8 October, but forks and a paid "ARTEX PRO" remain. The 13-page alert lists IOCs and advises rate limits on sensitive APIs and watching for bursts of automated requests. BleepingComputer (source)
🟡 Worth knowing

ARTEX follow-up: Malwarebytes adds a China-aligned user and defender advice

Added after publishingARTEXChina-alignedConnectWise credentials

WhatMalwarebytes adds that Microsoft reported a China-state-aligned group using ARTEX after compromising ConnectWise application credentials.

DoInventory forgotten employee and supplier portals, restrict admin interfaces, and scan for exposed passwords and API keys and rotate them.

Earlier coverage: 13/1112/1111/11
Details and sources
covering the FBI FLASH of 13 November (IC3 reference 261113), Malwarebytes adds that Microsoft separately reported a China-state-aligned group using ARTEX after compromising ConnectWise application credentials (Microsoft's report not read here). ARTEX's planning agent splits the work into tasks that other agents run with security tools, with little human input once started; targeted sectors include finance, government, healthcare and technology. Advice: patch internet-facing systems and inventory forgotten employee and supplier portals, restrict admin interfaces, scan for exposed passwords and API keys and rotate them, watch for request bursts, unusual data access and privilege changes, and prepare to revoke access and isolate systems quickly. Malwarebytes (source)

Law Enforcement 2

🟡 Worth knowing

Operation TorNado: 354 dark web vendors and buyers arrested

EuropolOperation TorNadodark web

WhatOperation TorNado led to 354 arrests of dark web vendors and buyers across 14 countries, with large drug, firearm and cash seizures.

Details and sources
Europol and police in 14 countries ran the operation from May 2025 to October 2026 against vendors and buyers from the closed Archetyp, Bohemia, Nemesis, Kingdom and Tor2Door markets, forums, single-vendor shops and encrypted chats. Arrests: 132 in the US, 59 in the UK, 45 in Hungary, 41 in Germany. Seizures include over 4.8 tonnes of drugs, 152 firearms and over €150 million in the US, plus 950 kg of drugs, 57 firearms and over €60 million in Europe. Help Net Security (source) · Europol (source)
🟡 Worth knowing

Philippines arrests four Taiwanese wanted for cyber fraud

PhilippinesTaiwancyber fraud

WhatPhilippine immigration arrested four Taiwanese wanted for online fraud and two Chinese fugitives tied to an illegal gambling ring.

Details and sources
the Bureau of Immigration's Fugitive Search Unit, acting on a request from Taiwan's office in Manila, arrested four Taiwanese accused of running an online fraud operation (New Taipei warrants of 20 August), plus two Chinese fugitives wanted over an illegal online gambling ring of 360 websites and a job-offer scam. All six await deportation. Manila Times (source)
‹ 13/11/202617/11/2026 ›