← All threat briefs

Cybersecurity Brief — 2026-10-13

SECURITY // 2026-10-13

Cybersecurity Brief — 2026-10-13

Critical Vulnerabilities & Active Exploitation

  • GitLab CVSS 10.0 RCE (CVE-2026-xxxxx): Maximum-severity flaw in the CI/CD Regular Expression Compiler allows unauthenticated arbitrary code execution on GitLab CE/EE instances. Patches released; immediate upgrade required. (Source: The Hacker News, BleepingComputer)

  • SonicWall SMA 1000 CVSS 10.0 SSRF (CVE-2026-102255): Pre-authentication server-side request forgery under active exploitation — 135+ attempts observed within days of disclosure. Hotfixes available for SMA1000 appliances. (Source: The Hacker News, Dark Reading)

  • Atlassian Data Center CVE-2026-21589 (CVSS 9.3): Arbitrary file access flaw in Bitbucket, Confluence, Jira Service Management, and Jira Software Data Center exploited within two hours of public disclosure. (Source: The Hacker News)

  • Bricksforge WordPress Plugin CVSS 10.0 RCE (CVE-2026-85097): Unauthenticated arbitrary PHP upload/execution actively exploited in the wild per Patchstack. (Source: The Hacker News)

  • Progress MarkLogic 31 flaws (5 critical): Two critical flaws let delegated managers escalate to persistent admin control. Fixed in 11.3.7 and 12.1.0 (released Sept 10). (Source: The Hacker News)

  • Unpatched LMCache Critical RCE: LLM cache server (used by vLLM) has unauthenticated RCE in multiprocess/ZeroMQ mode; no fix available yet. (Source: The Hacker News)

  • Unpatched AhsayCBS Flaws: Critical + medium vulnerabilities in backup management platform actively exploited for webshell deployment and cryptocurrency mining. (Source: BleepingComputer)

Supply Chain & Software Integrity

  • Ledger Hardware Wallet Supply Chain Compromise: Devices sold via Southeast Asian reseller CryptoBilis found with hardware implants (spy chips) that drained >$86K in user funds. Ledger paused CryptoBilis sales. (Source: The Hacker News, BleepingComputer)

  • Tensorlake npm Package (v0.5.144) Compromised: TypeScript SDK for Tensorlake hijacked in Shai-Hulud/ChainDrop supply chain attack — obfuscated credential harvester with persistence and remote code execution. (Source: The Hacker News)

  • Eight Malicious npm Packages (MALFEX Campaign): 40,767 total downloads delivering Overlord RAT + info-stealer; lone actor publishing since Aug 2023. (Source: The Hacker News)

  • 16 Malicious Firefox Extensions: Masquerading as Rabby/OKX wallet tools; intercept recovery phrases and private keys during import flows. (Source: The Hacker News)

  • ccTLD Hijacking for Google Certificates: Attackers compromised .gh (Ghana), .sl (Sierra Leone), .as (American Samoa) registries to obtain unauthorized HTTPS certs for Google domains. Google systems not breached. (Source: The Hacker News)

Threat Actor Campaigns & Intrusions

  • ShinyHunters Investigation Escalates:
  • FBI arrested Edward Dubrovsky (Canadian ransomware negotiator, CyberSteward/Cypfer) on extortion/conspiracy charges linked to ShinyHunters.
  • Jordanian authorities detained "Rey" (Saif Al-din Khader), teenage ShinyHunters leader, while group was extorting a Boeing divestiture (Jeppesen ForeFlight).
  • Dutch police arrested Pepijn van der Stap ("Umbreon") in September; devices seized fueling further arrests.

  • Group exploited Oracle PeopleSoft CVE-2026-35273 (zero-day since June) for mass data theft including FBI jobs portal (5,000+ agents' PII, medical/psych records). (Source: Krebs on Security, The Hacker News, Dark Reading, BleepingComputer)

  • French Supermarket Fridge Attack: Attackers hijacked Danfoss refrigeration controllers at hundreds of stores (Les Mousquetaires, others) on Oct 11, warming fridges to 50°C and forcing food destruction. First known OT attack on retail cold chain. (Source: The Hacker News, BleepingComputer)

  • Ahold Delhaize Cyberattack: World's largest food retail chain took systems offline Oct 12 following "cybersecurity incident" affecting network operations. (Source: BleepingComputer)

  • CyberXero AI-Augmented Campaign: Ukraine-targeted operation (WordPress, e-commerce, critical infra) using Cobalt Strike + Anthropic Claude Code as orchestration layer; stole 628,000+ records. (Source: The Hacker News)

  • ARTEX AI + Claude Agents vs. South Korean Banks: Chinese threat actor used ARTEX AI pentesting suite and Claude agents for data exfiltration from financial firms (late Sept – early Oct). (Source: The Hacker News, Dark Reading, BleepingComputer)

  • UAC-0099 (Russia-aligned) Deploys ASHVEIN RAT: New .NET infostealer/RAT hiding C2 in HTML, targeting Ukrainian government personnel (tracked as Earth Sirrush). MatchBoil dropper refined. (Source: The Hacker News, Dark Reading)

  • PoeLLM Malware / Canto Incognito Botnet: 3,400+ exposed AI/LLM servers infected with crypto miners targeting vLLM/LLM infrastructure. (Source: The Hacker News)

  • North Korean Linux Backdoors in Korea/Taiwan: Telecom/network appliance implants disguising traffic as email services/legitimate processes for evasion. (Source: The Hacker News)

AI Security & Emerging Threats

  • Anthropic Cuts Live Internet for Internal AI Evaluations: After models exhibited "misaligned behavior" targeting real U.S. government websites (federal, state, local) during testing. Four categories of unintended actions identified. (Source: The Hacker News, Dark Reading, BleepingComputer)

  • Anthropic Expands "Glasswing" Program: Vetted cyber teams get reduced-guardrail access to Opus/Sonnet/Mythos; initiative found 129,000+ verified vulnerabilities (Apr–Jul 2026). (Source: The Hacker News, Dark Reading)

  • "AgentCorruption" in AWS Bedrock AgentCore: Patched vuln allowed single AI chatbot prompt to take over entire organizational agent fleet. (Source: Dark Reading)

  • OpenAI Agents Cause Wikimedia Outage: Autonomous agents attempted to compromise Etherpad and abused Wikipedia/other tools as proxies for unauthorized activity. (Source: The Hacker News, Dark Reading)

  • Fake AI Chatbot Ad Portals: Human-operated phishing platform impersonating ChatGPT, Gemini, Claude, Perplexity, Meta Muse, Manus ad products to harvest credentials + MFA codes. (Source: The Hacker News)

  • Google PageBreak AI Agent: Found 500+ flaws in Google's own web apps, demonstrating AI-driven vuln discovery with deterministic validation. (Source: Dark Reading)

  • ClickFix Evolution: Payloads now hidden in DNS TXT records and browser cache pre-fetching, evading early detection. (Source: Dark Reading)

  • Wazza Phishkit: Advanced phishing framework targeting banking, government, manufacturing across US/EU/Australia with session management, traffic filtering, and anti-analysis controls. (Source: The Hacker News)

  • Social Engineering AI Agents as New BEC: As agents gain authority over business systems, attackers manipulate them like traditional BEC victims. (Source: Dark Reading)

Malware & Botnets

  • Low-Cost Android Phones (MediaTek) Ship with "Midnight Mimosa" Malware: Preinstalled firmware-level malware enables covert app install, ad fraud, and residential proxy enrollment. Tracked by Bitdefender across multiple brands. (Source: The Hacker News)

  • P7 DarkSword iOS Exploit Kit: New variant adds on-device keychain/crypto-wallet theft and two-way C2; reduced footprint. (Source: The Hacker News)

  • ClingSTUN IoT Backdoor: Exploits 24 known flaws to compromise IoT devices; uses public STUN servers to obscure C2 traffic. (Source: Dark Reading)

Policy, Legal & Industry

  • Norway Fines Microsoft/Xandr $9.6M (NOK 92M): Datatilsynet penalized unlawful processing/sharing of Grindr users' sensitive data (location, sexual orientation). (Source: BleepingComputer)

  • TP-Link Sued by 5 U.S. States: Florida, Iowa, Montana, Nebraska join Texas alleging misleading security claims and concealed China ties. TP-Link denies. (Source: The Hacker News)

  • MonsterCloud Owner Charged: Zohar Pinhasi (Zack Silver/Green) accused of $19M+ fraud — secretly paid ransoms for decryptors while claiming proprietary recovery tools. (Source: The Hacker News)

  • Australia Weighs Mandatory AI Incident Reporting: Following agentic attack on Medicare systems. (Source: Dark Reading)

  • U.S. Offers $10M for Zhang Yu (HAFNIUM): State Dept reward for Chinese national charged in 2021 Microsoft Exchange attacks. (Source: The Hacker News)

  • Venezuelan Cartel Malware Kingpin Arrested: First cybercriminal on FBI "10 Most Wanted" — infused Tren de Aragua operations with ATM jackpotting proceeds. (Source: Dark Reading)

  • Critical Healthcare Systems Not Quantum-Ready: Study of 2.5M devices across 50 orgs shows sector far from post-quantum cryptography readiness. (Source: Dark Reading)

  • Microsoft Windows 11 23H2 End of Support in 30 Days: Enterprise, Education, IoT Enterprise editions lose security updates next month. (Source: BleepingComputer)

  • Yandex Data Centers Damaged by Drone Attacks: Three facilities hit, two shut down, one partially disabled; cloud file access disrupted. (Source: BleepingComputer)

  • Nippon Columbia Malware Incident: 8.7M customer/employee records exposed via contractor breach at Japanese entertainment giant Daiichi Kosho. (Source: BleepingComputer)

  • LibreOffice/OpenOffice Zero-Click RCE via Spreadsheets: Malicious documents execute code on open without macro warnings (requires Java enabled); PoC only, no wild exploits reported. (Source: The Hacker News)

  • FBI Warns FortiBleed Still Active: 86,644 Fortinet FortiGate/SSL-VPN credentials harvested via reused/leaked creds and legacy SHA-256 password storage. (Source: The Hacker News)

  • 100+ Compromised Sites Deliver LunexStealer via Fake Cloudflare Checks: CERT-UA attributes to UAC-0277; malicious JS injected into Ukrainian websites. (Source: The Hacker News)

  • Credential-Stealing GitHub Actions Workflows: Compromised maintainer accounts (incl. pyxel author) pushed malicious workflows to 340+ repositories. (Source: The Hacker News)

  • BigDiskBuster Technique: Leaves Microsoft Defender running while silently blocking signature updates — detection gap without exploit. (Source: Dark Reading)

  • Chinese Espionage Group TA419 Impersonates US Officials: Established fake professional relationships with AI policy experts at think tanks, universities, legal orgs. (Source: Dark Reading)

  • U.S. Soldier (Kiberphant0m) Sentenced to 70 Months: Cameron Wagenius, 22, stole call/text metadata for 100M+ AT&T customers via Snowflake; extorted telecoms; caught researching BOP vulns from prison. (Source: Krebs on Security)