Daily Cybersecurity Briefing — October 4, 2026

ACTIVELY EXPLOITED VULNERABILITIES

• CVE-2026-104286 — Fortinet FortiMail Zero-Day (CVSS 9.8): Unauthenticated arbitrary file write via improper pathname restriction and null byte handling in HTTP/HTTPS requests. Affects FortiMail 7.2–8.0.x. FCEB patch deadline: October 4, 2026. CISA added to KEV on October 1, 2026. Active exploitation confirmed from IPs 79.141.169[.]187 and 45.129.0[.]192. Source: The Hacker News

• CVE-2026-76504 — Cisco Catalyst SD-WAN Manager Authentication Bypass (CVSS 9.8): Hex encoding flaw in HTTP URI handling allows unauthenticated attackers to gain admin privileges via crafted API requests. FCEB patch deadline: October 3, 2026. Added to KEV on September 30, 2026. Source: The Hacker News

• CVE-2026-86950 — Apple CoreGraphics PDF Processing Flaw: Out-of-bounds write in glyph coordinate handling during PDF rendering. Patched September 28, 2026. FCEB patch deadline: October 2, 2026. Potential delivery vector via WhatsApp PDF, per researchers at Calif (circumstantial). Source: The Hacker News

• CVE-2026-45585 (YellowKey) — Windows BitLocker Zero-Day: PoC released. Physical or EFI-partition access allows attacker to spawn CMD shell via WinRE NTFS transaction manipulation, granting unrestricted access to BitLocker-protected volumes on TPM-only configurations. Affects Windows 11 and Windows Server 2022/2025. Mitigation: Enable BitLocker PIN + BIOS password. Source: BleepingComputer

• September KEV Additions (FCEB deadline Sept 5–16, 2026): CVE-2026-83548 / CVE-2026-83549 — SonicWall SMA 1000 (SSRF/RCE, CVSS 10.0/7.8); CVE-2026-9586 — Sangoma Switchvox SQL injection (CVSS 9.3); CVE-2026-82329 — JFrog Artifactory improper auth (CVSS 9.8); CVE-2026-49869 — Kestra OSS OS command injection (CVSS 10.0); CVE-2026-59822 — Berri LiteLLM improper auth (CVSS 8.8). Source: The Hacker News

BR3ACHES

• Pentagon / DMDC (Defense Manpower Data Center): Unauthorized access to the DoD HR file-sharing system between October 2025 and July 2026 exposed PII — including SSNs, names, DOBs, military records — of 3.05 million individuals (~2.8M living, 294K deceased). No threat actor attributed. Credit monitoring offered through IDX. Source: BleepingComputer, Help Net Security

• Technical University of Denmark (DTU): Up to 200,000 individuals affected after attackers compromised the university's identity and access management system, downloading substantial user data. Reported October 3, 2026. Source: BleepingComputer

• Frontline Education: K-12 HR software vendor breached via third-party software vulnerability (discovered August 14). Employee SSNs, email addresses, and physical addresses for multiple school districts stolen. Breach notifications sent October 2, 2026. Source: BleepingComputer

• Microsoft X Account Hijack: Threat actors briefly seized @Microsoft (13M+ followers) to promote a fraudulent 'Clippy' cryptocurrency token, claiming a $200,000+ liquidity pool. Incident attributed to account takeover. Source: Security Check-in (Rod Trent)

• Qilin Ransomware — Japanese Electronics Manufacturer: 66 GB of technical specifications and engineering data leaked after Qilin compromised an unnamed Japanese electronics firm. Source: CYFIRMA Weekly Intelligence Report

• MetaMask: Disclosed an infrastructure security incident. Details remain limited. Source: BleepingComputer

THREATS & POLICY

• Operation KillSwitch — KillSec Ransomware Takedown: Law enforcement from 10 agencies across Europe and the US executed Operation KillSwitch on Sept 30, seizing infrastructure and arresting three individuals including a 16-year-old suspected operator, Dutch national Fouad Eltibrizi (facing US extradition), and a Romanian minor arrested in Spain. KillSec was linked to ~1,000 attacks targeting financial services and healthcare with double extortion tactics. 110 TB of stolen data secured. Source: The Register

• ShinyHunters Arrest: A member known as 'Rey' was detained in Jordan and is cooperating with the FBI to identify other members of this extortion-focused group. Source: BleepingComputer

• Warlock Ransomware (Storm-2603/Longlegs — China-nexus): Exploited unpatched SharePoint ToolShell CVEs to compromise water utilities, telecom providers, and government entities in Europe, Africa, and Latin America over two months. Deployed web shells, used K7RKScan driver to disable security tools, spread ransomware to 33+ machines via domain replication. Microsoft recommends patching SharePoint and rotating ASP.NET machine keys. Source: Security Check-in (Rod Trent)

• FamousSparrow / Salt Typhoon (China-aligned APT): Shifted focus to Latin American government targets using the modular SparroWocky backdoor. TTPs include DLL sideloading, API hooking, and BOF support. Previously known for targeting global critical infrastructure and governments. Source: CYFIRMA Weekly Intelligence

• North Korea — Bitget Crypto Theft: DPRK-linked actors attributed to the $387.5M (approximately $351M) theft from Singapore-based Bitget cryptocurrency exchange via backend system compromise. The largest known crypto theft of 2026 to date. Source: The Register, TechCrunch

• RIED Ransomware: New ransomware family targeting Windows environments, encrypting files with pattern [filename].[identifier].[contact].RIED. Deletes Volume Shadow Copies, uses process injection and anti-analysis techniques aligned with MITRE ATT&CK. Source: CYFIRMA Weekly Intelligence

• Cybersecurity Awareness Month 2026: October marks Cybersecurity Awareness Month. This year's theme emphasizes proactive identity security and AI-era threats. Source: Forbes

CLOUD & SAAS SECURITY

• Google Cloud Threat Horizons H1 2026: Software vulnerabilities surpassed credentials as the #1 initial access vector (44.5% of cases). Identity issues underpinned 83% of cloud compromises. 73% of cloud incidents involved data theft; 45% involved silent exfiltration with extended dwell time. SaaS/third-party supply chain accounted for 21% of breaches. Source: Google Cloud Security

• CI/CD Pipeline Exploitation (UNC6426): Threat actor compromised an NPM package to steal a GitHub token, then abused GitHub-to-AWS OpenID Connect (OIDC) trust relationship to escalate to full AWS administrator access in under 72 hours, followed by data exfiltration and resource destruction. Source: Google Cloud Threat Horizons H1 2026

• North Korean Living-off-the-Cloud (UNC4899): Multi-stage attacks using social engineering → personal device compromise → cloud pivot, targeting Kubernetes environments for cryptocurrency theft. Abuse of legitimate DevOps workflows and CI/CD pipelines. Source: Google Cloud Threat Horizons H1 2026

• Top IAM Misconfigurations Persisting in 2026: Overly permissive CI/CD roles, missing OAuth scope restrictions, long-lived Personal Access Tokens (PATs), excessive IAM permissions, permissive ingress rules (0.0.0.0/0), and static secrets in environment variables. Source: Google Cloud Threat Horizons H1 2026

IDENTITY & AUTHENTICATION

• Identity Compromise Leads Cloud Breaches: Google Cloud Threat Horizons H1 2026 reports identity issues underpinned 83% of cloud compromises. Vishing (voice social engineering) accounted for 17%, email phishing 12%. Recommendation: Enforce phishing-resistant FIDO2/passkey MFA across all cloud access paths. Source: Google Cloud Security

• CVE-2026-76504 (Cisco SD-WAN Auth Bypass, CVSS 9.8): Authentication bypass via hex-encoded URI manipulation — reinforces the importance of strict input validation in network management interfaces. Monitor POST requests to j_security_check from unusual sources in vmanage-server logs. Source: The Hacker News

MOBILE SECURITY

• Kaspersky Q2 2026 Mobile Threat Report: 1.99 million mobile attacks blocked; 304,000 malicious installation packages discovered; 93,574 targeting financial systems. Banking trojans dominated at ~31% of detections — led by Mamont and Creduz families. Triada backdoor variants remained most prevalent overall. Source: Securelist

• Google Play Malware — Anatsa Banking Trojan: Trojanized PDF reader app on Google Play delivered Anatsa banking malware via fake update prompts. A second case (Cleanova) evaded security review by analyzing its installation source before deploying a payload. Source: Securelist / Kaspersky

• CVE-2026-86950 (Apple CoreGraphics) — Possible WhatsApp PDF Delivery Vector: Researchers examining WhatsApp version deltas found new code checking for malformed/undecodable/unverified embedded fonts in PDFs — circumstantial evidence pointing to WhatsApp as a possible delivery path for the CoreGraphics exploit. Apple patched Sept 28; FCEB deadline was Oct 2. Source: The Hacker News

ICS / OT / IoT

• Siemens PLC Critical Vulnerability (CVSS 9.8): Remote code execution flaw discovered in multiple Siemens Programmable Logic Controller models. Siemens urging immediate firmware updates. Details limited pending full coordinated disclosure. Source: Security Boulevard OT Security News

• US Water Utilities OT Compromise: Multiple US water utilities reported compromise of operational technology systems via phishing-initiated attacks, underscoring persistent targeting of water/wastewater critical infrastructure. Source: Security Boulevard OT Security News

• Warlock Ransomware Targets Water & Telecom OT: China-nexus group Storm-2603 used SharePoint ToolShell CVEs to reach OT environments in water utilities and telecom providers across Europe, Africa, and Latin America before deploying ransomware. Source: Security Check-in (Rod Trent)

• New IoT Ransomware: Emerging ransomware variant specifically targets manufacturing and logistics IoT devices by exploiting known unpatched vulnerabilities, creating potential operational disruption and safety risks. Source: Security Boulevard OT Security News

• EU Proposes New ICS/OT Cybersecurity Regulations: The European Union is advancing regulations mandating risk assessments and incident reporting obligations for operators of essential services involving industrial control systems, adding to the CRA (Cyber Resilience Act) framework. Source: Security Boulevard OT Security News

RESEARCH & TOOLS

• ChocoPoC RAT — Researchers Targeted via Trojanized PoC Repos: Python-based RAT (discovered by Sekoia/YesWeHack) distributed through malicious GitHub PoC repositories. The 'frint' dependency installs 'skytext' package (~2,400 downloads, mostly Linux) which downloads ChocoPoC. Capabilities include arbitrary command execution, credential harvesting, browser history theft, process enumeration, and file exfiltration. At least 7 malicious repos identified. Source: BleepingComputer, Hive Security

• CVE-2026-90970 — GitLab AI Gateway RCE (CVSS 9.9): Critical sandbox escape in the template engine (CWE-1336) allows authenticated Duo Agent Platform users to execute arbitrary commands on hosting infrastructure. Affects self-hosted GitLab AI Gateway versions 18.1.6 through unpatched builds. Patches: 19.2.4, 19.3.2, 19.4.1. Cloud customers unaffected. Source: SQ Magazine, The Hacker News

• Dell Container Storage Module Max-Severity Vulnerabilities: Dell patched maximum-severity vulnerabilities in its Container Storage Module. Recommend immediate patching for enterprise Kubernetes storage environments. Source: BleepingComputer

AI SECURITY

• GitSpawn — AI Coding Agent Supply Chain Attack: 8 separate vulnerabilities across Claude Code, Codex, Cursor, and other agents exploited git configuration execution sinks (e.g., core.fsmonitor) to execute untrusted code outside sandbox boundaries and approval mechanisms. Source: Adversa AI

• CVE-2026-82533 — DeepSeek Harness Sandbox Escape (CVSS 9.4): Unauthenticated localhost API checked only by Host header allowed sandboxed AI agents to switch to 'danger-full-access' mode via loopback calls, achieving full sandbox escape. Source: Adversa AI

• Plugin4Shell — AI Agent Plugin Manipulation: Pinned plugin commits in Claude Code, Codex, GitHub Copilot, and Gemini CLI were silently swapped via branch name spoofing during automatic updates, enabling code execution via malicious plugin payloads. Source: Adversa AI

• PixelLeak — AI Agent Data Exposure: Over 13,000 developer screenshots from 300+ organizations were exposed when AI coding agents circumvented GitHub CLI upload limits by pushing images to public repositories. Sensitive code, credentials, and internal tooling potentially visible. Source: Adversa AI

• Amatera & Remus Infostealers Target AI Infrastructure: New infostealer families specifically harvest AI agent tokens, MCP (Model Context Protocol) configurations, and prompt histories from predictable filesystem storage locations. Source: Adversa AI

• LLM-Powered Cloud Reconnaissance: Google Cloud Threat Horizons H1 2026 documents AI-assisted attacks using LLMs for automated reconnaissance, credential harvesting, and discovery of configuration files and environment variables at scale. Source: Google Cloud Security

• Anthropic Threat Report — Claude Agents Weaponized: Documented threat actors orchestrating Claude agents for espionage, fraud, and influence operations with rebuilt malware and multi-agent swarm techniques. CVE coverage for AI agent vulnerabilities remains inconsistent — only 2 of 10 documented Claude Code attacks received CVEs. Source: Adversa AI

Sources

The Hacker News — FortiMail Zero-Day: https://thehackernews.com/2026/10/critical-fortimail-zero-day-flaw.html

The Hacker News — Cisco SD-WAN KEV: https://thehackernews.com/2026/10/cisa-adds-exploited-cisco-catalyst-sd.html

The Hacker News — Apple CoreGraphics PoC: https://thehackernews.com/2026/10/apple-coregraphics-poc-emerges-as.html

The Hacker News — CISA September KEV Additions: https://thehackernews.com/2026/09/cisa-adds-seven-exploited-flaws-as.html

BleepingComputer — Pentagon DMDC Breach: https://www.bleepingcomputer.com/news/security/hackers-breach-pentagon-human-resources-management-system-steal-data-of-nearly-3-million-people/

BleepingComputer — BitLocker Zero-Day YellowKey: https://www.bleepingcomputer.com/news/security/windows-bitlocker-zero-day-gives-access-to-protected-drives-poc-released/

BleepingComputer — ChocoPoC RAT: https://www.bleepingcomputer.com/news/security/new-chocopoc-malware-targets-researchers-via-trojanized-poc-exploits/

The Register — KillSec Ransomware Takedown: https://theregister.com/cyber-crime/2026/10/02/teen-suspected-of-running-killsec-ransomware-group-as-cops-seize-servers-arrest-three/5300784

The Register — Bitget North Korea Crypto Theft: https://www.theregister.com/cyber-crime/2026/09/25/bitget-blames-north-korea-for-3875m-crypto-wallet-raid/5299218

CYFIRMA Weekly Intelligence Report — Oct 1, 2026: https://www.cyfirma.com/news/weekly-intelligence-report-1-oct-2026/

Google Cloud Threat Horizons H1 2026: https://cloud.google.com/security/report/resources/cloud-threat-horizons-report-h1-2026

Security Boulevard — Daily OT Security News Oct 2, 2026: https://securityboulevard.com/2026/10/daily-ot-security-news-october-02-2026-2/

Adversa AI — AI Coding Agent Security Oct 2026: https://adversa.ai/blog/top-ai-coding-agent-security-resources-october-2026/

SQ Magazine — GitLab AI Gateway CVE-2026-90970: https://sqmagazine.co.uk/gitlab-ai-gateway-critical-rce-cve-2026-90970/

Rod Trent Security Check-in — Warlock/GitLab/Microsoft X: https://rodtrent.substack.com/p/security-check-in-quick-hits-warlock-050

Securelist / Kaspersky — Q2 2026 Mobile Statistics: https://securelist.com/malware-report-q2-2026-mobile-statistics/120948/