HeadFlash

Security

Check Point zero-day exploited to root firewalls, wiper hits ESXi

Check Point rushed hotfixes for a root-level zero-day exploited against dozens of customers, plus Turla's Exchange malware and a 1.3M-device Android backdoor.

Listen

This edition was produced with artificial intelligence. Text and voice are generated automatically.

Check Point issues emergency hotfixes for exploited root zero-day

Check Point released emergency hotfixes for a high-severity local privilege escalation flaw, CVE-2026-123039, exploited since late October to gain root and establish persistence. The bug sits in the DynamicCLI IPC interface of Gaia OS: the root-level dynamic_server daemon listens on a world-writable local socket without authenticating connecting processes, so any authenticated local user can have commands executed as root. Check Point observed active exploitation against a couple of dozen customers. Attackers first breached the Mobile Access portal via the PHP2Shell vulnerability, then escalated to root, dropped webshells and installed TSGA backdoor and TSProxy malware, pivoted to a VMware ESXi host and deployed a destructive wiper. Firewall, Security Gateway, management servers and Scalable Platforms are affected. Teams unable to patch immediately should stop dynamic_server and prevent it from restarting.

Check Point warns of zero-day flaw exploited for root access →

Topics: Zero-day vulnerabilitiesPatch Tuesday & security updates

Turla breached Romanian government Exchange servers to read NATO and EU email

ESET attributed a breach of Romanian government Microsoft Exchange servers to Turla, the Russian FSB-linked espionage group active since at least 2004. The toolkit, called Cereal, monitors email traffic, steals selected messages and hides inside legitimate Exchange processes, leaving little on disk and using no external command-and-control infrastructure. Instead, the compromised server itself served as the C&C hub, a technique ESET says only a handful of APT groups use and not previously documented in other malware families. ESET found the toolset late last year at a government organization in Kazakhstan and soon after in Romania, compromised on December 23, 2025, configured to flag emails sent to or received from NATO and the EU. It remains unclear how attackers got initial access or how much data was taken.

Russian hackers breach Romanian government email servers | Cybernews →

Topics: State-sponsored hacking

Google ties 1.3 million infected Android devices to Videostrong backdoor

Google’s Threat Intelligence Group concluded with high confidence that China-based manufacturer Videostrong deliberately built a backdoor called XCPE into Android devices and operated it for profit, affecting more than 1.3 million devices worldwide, mostly Android TV streaming boxes plus karaoke machines, digital signage and some telehealth hardware. Videostrong sold access to criminal affiliates who installed more malware and enrolled devices in a botnet used for residential proxy traffic and fake ad activity; operators also captured screenshots, injected keystrokes and collected account data. XCPE was concealed during Google’s testing, loaded after first startup, could disable Play Protect and reinstalled itself after reboot, so a factory reset might not clean it. Google took legal action to shut down command-and-control domains, banned Videostrong from certified Android devices and strengthened firmware verification, but disrupting servers does not remove malicious firmware already in homes.

Android TV box malware found in 1.3M devices →

Topics: Malware & infostealers

Critical Apache HTTP Server flaw exposes 64% of cloud environments

A critical race condition, CVE-2026-119618, in Apache HTTP Server’s mod_http2 module allows a remote, unauthenticated attacker to corrupt heap memory and achieve arbitrary code execution with the web server’s privileges. The flaw, in the assign_headers() function, was independently discovered by Anthropic and Google Cloud’s Vulnerability Research Team. No exploit is available and no in-the-wild exploitation has been reported, but Wiz data shows 64% of cloud environments have vulnerable resources and 51% have publicly exposed instances. Affected versions are 2.4.55 through 2.4.70, patched in 2.4.71, and standalone mod_h2 2.0.8 through 2.0.42, patched in 2.0.42. Administrators who cannot patch immediately should disable HTTP/2 and restrict active protocols to HTTP/1.1.

Critical RCE Vulnerability in Apache HTTP Server: Everything You Need To Know | Wiz Blog →

Topics: Zero-day vulnerabilities

NvRoot GPU driver flaw lets unprivileged users take over rented GPU hosts

Stealthium researcher Branislav Brzak built a working exploit for NvRoot, CVE-2026-47591, a local privilege escalation in NVIDIA GPU drivers that yields root. The driver’s memory-pinning path enforces CPU and GPU permission fields separately, but vulnerable drivers derive page permission from the CPU field alone; a request marking CPU read-only and GPU writable pins a read-only file’s shared cache while the GPU mapping stays writable. An unprivileged user can pin a setuid binary such as /usr/bin/su, overwrite its in-memory copy via a GPU write and execute it, with the on-disk file unchanged. The bug is deterministic, kernel-version independent and works from an unprivileged container, making it a host-level compromise for every tenant on a rented GPU. NVIDIA rates it 7.8 High; Stealthium argues for Critical in multi-tenant environments. Fixed versions include 610.57.04, 595.91.07, 580.178.04 and later; no workaround exists.

From Rent to Root: Introducing NvRoot | Stealthium →

Topics: Nvidia

Fake OpenAI and SpaceX event pages steal Microsoft and Google sessions

Hackers are luring employees with fake OpenAI, SpaceX and Anthropic event invitations that steal Microsoft and Google credentials and authentication sessions, according to Push Security. The registration pages promise free premium AI access, such as three months of ChatGPT Pro or Grok Heavy, and require signing in with a corporate account. Two phishing-as-a-service platforms do the work: Kali365 runs Microsoft device code phishing, while VoidProxy acts as an adversary-in-the-middle reverse proxy to intercept Google sign-ins and session cookies, and serves as a second stage for Microsoft. Both bypass multifactor authentication. Push Security blocked visits to the pages in late October 2026, finding victims redirected through Google and Zapier URL shorteners, with two kits tracked as ZoomMock and MSDRoot. Defenses include restricting device code authentication and using phishing-resistant FIDO2 keys.

Fake OpenAI, SpaceX events used to steal Microsoft, Google accounts →

Topics: Phishing

Datadog tracks ARTEX AI penetration framework hitting AWS environments

Datadog observed in-the-wild activity in October 2026 from ARTEX, an AI-native penetration framework, against several AWS environments, focused on financial data. Using compromised long-term IAM access keys, ARTEX searched CloudWatch via FilterLogEvents for financial information, file paths and secrets including long-lived AKIA keys and passwords; it also abused public Amazon Cognito user pools with anonymous sign-ups, token collection and AWS account enumeration. Datadog tied the activity to IPs resolving to a hosted console and found hundreds of additional internet-exposed ARTEX consoles, often in Mandarin and hosted in locations including Singapore and Hong Kong. A Chinese developer published ARTEX on GitHub in July 2026 and made it closed-source in October after it passed 12,000 forks; Microsoft tied it to China-aligned actor Storm-0940. Hunting should start with ARTEX user agents, known IPs and spikes in permission-denied errors.

On the hunt for ARTEX: understanding agentic attacks in AWS environments | Datadog Security Labs →

Topics: Autonomous AI agent cyberattacks in the wild

SignDrop Windows zero-day turns Defender into a path to SYSTEM

Abdelhamid Naceri, a former Microsoft employee publishing unpatched Windows bugs as Nightmare Eclipse, released a zero-day exploit called SignDrop that lets a standard user gain SYSTEM-level control of a PC. The flaw is in the Common Log File System: each log creates a companion recovery file that inherits its permissions from the log, so whoever creates the log controls the companion. SignDrop tricks Microsoft Defender, which runs with the highest privileges, into opening an attacker-prepared log, creating the companion file with Defender’s rights while the attacker keeps control, reaching a SYSTEM command prompt. The attack subverts Microsoft’s log-signing safeguard and requires Defender to be running. Analyst Will Dormann reproduced it only on Windows 11 26H2 machines set up more than 90 days. Microsoft says it is investigating, has seen no in-the-wild exploitation and claims Defender blocks the attack; a fix likely will not arrive before December’s Patch Tuesday.

New Windows zero-day turns Microsoft’s own security fix into a path to full system access →

Topics: Windows securityZero-day vulnerabilities

PostgreSQL patches 52 flaws, including a two-decade-old code execution bug

The PostgreSQL Global Development Group released security updates on November 12, 2026: versions 18.7, 17.12, 16.16, 15.20 and 14.25, the final release of the 14 branch. The batch fixes 52 vulnerabilities. CVE-2026-6460, a type confusion bug in the core server rated 8.8 High, lets a user who can create database objects run arbitrary code as the operating system user running the database, primarily via ALTER … SET DATA TYPE. Press coverage says the flaw traces back more than 20 years and affects every supported major version, 14 through 18; it is not known to be exploited and there is no workaround for servers that cannot be updated. Other fixes include CVE-2026-6467, where extension script schema mishandling can lead to arbitrary SQL execution. Minor releases are cumulative, so administrators can swap binaries and restart without a dump and reload.

PostgreSQL Patches 52 Vulnerabilities, Including a Code Execution Flaw That Reportedly Dates Back Two Decades - Nationwide Private Investigator & Skip Tracing | Knoxville TN Background Checks & OSINT | Kyle's Investigation →

Topics: Patch Tuesday & security updates

Cairn AI framework ran 25 autonomous attacks on chip and defense targets

From August to September 2026, Chinese-speaking attackers used an open-source AI penetration-testing framework called Cairn, paired with Chinese models Zhipu GLM and DeepSeek, to autonomously conduct 25 operations against at least 19 organizations. Attackers supplied a target and a one-line goal; Cairn planned, executed and recorded results on its own. The campaign achieved unauthorized access or data retrieval at 10 organizations, including dozens of valid credentials, exposure of more than one million personal records at one organization, creation of persistent administrator accounts in a production system and modification of a live industry standards-ballot document. Unit 42 assesses with moderate confidence the activity represents China-nexus intelligence collection tracked as CL-STA-1238. Targets concentrated in semiconductor and advanced-manufacturing supply chains, defense and defense policy, and human rights and civil society. Exploited vulnerabilities were known: authentication bypasses, weak and default credentials, exposed cloud APIs and dated CVEs.

Attackers Use Autonomous AI to Target Semiconductor, Defense and Civil Society Sectors →

Topics: Autonomous AI agent cyberattacks in the wildState-sponsored hacking

FBI flags jackpotting surge as ATM malware thefts top $20 million

Two Wisconsin ATMs were robbed of hundreds of thousands of dollars in a jackpotting scam the FBI calls a new way to rob a bank. Video from a Glendale bank in March shows two men using a key to get inside an ATM, the screen going black and rebooting, and the men returning about an hour later to pull out stacks of cash, posing for a selfie; they got away with roughly $110,000. A week later, roughly $230,000 was stolen from a bank ATM in Greenfield, and it is unclear if the crimes were connected. FBI Milwaukee Special Agent in Charge Alan Karr said the robbers install malware and hack ATMs to dispense cash, calling it the new bank robbery without the violence. Banks have reported more than 1,900 jackpotting incidents since 2020; last year alone, 700 incidents cost banks more than $20 million. Karr urged banks to update software and increase physical protections.

FBI warns of growing bank robbery trend where thieves install malware on ATMs →

Topics: Malware & infostealers

Daily tech-news flash

The flash, every weekday.

Five minutes on AI, privacy and security — one short email per niche you pick, with a podcast to match.

Your niches