FlagThis — Daily Cybersecurity Intelligence Briefing

FILTERING BY: CLEAR FILTER

Trezor Safe 5 Hardware Implants Exfiltrate Recovery Phrases via Cellular Networks

Trezor confirmed three Safe 5 hardware wallets distributed by Southeast Asian reseller CryptoBilis contain post-manufacturing circuit board implants. The malicious PCBs tap mainboard power and data lines to intercept BIP-39 recovery phrases during initialization, then exfiltrate mnemonics via embedded GSM/LTE cellular modems using SMS, USSD, or mobile data. The supply-chain compromise occurred at the distribution layer, not the manufacturing facility. Trezor released firmware tampering detection and directed affected customers to migrate funds to new wallets with fresh recovery phrases.

Meta Muse AI Agent: Chained Prompt Injection to Cloud Takeover Patched

Meta Muse AI Agent, a personal AI agent with deep cloud integrations, was patched after researchers disclosed a chained prompt injection vulnerability leading to full cloud account takeover. Attackers delivered indirect prompt injection payloads via compromised web pages and browser extension content scripts, forcing the LLM to execute unauthorized tool calls that exposed cloud credentials, IAM tokens, and API keys from agent memory. Attackers then chained these credentials with Chromium renderer and kernel escape vectors to bypass agent sandboxing and pivot into the cloud control plane, enabling lateral movement across connected SaaS environments and persistence via modified agent workflows. Remediation included system prompt hardening, tool-use authorization boundaries, credential isolation, WAF rules, and LLM gateway filters.

Barracuda Cloud Breach Linked to ARTEX AI-Powered Cyberattack

Barracuda Networks disclosed a cloud account compromise affecting approximately 10 accounts over a 44-hour window. Attackers exploited an unretired cloud access key issued to a third-party AI vendor to infiltrate Barracuda's cloud environment. The intrusion is associated with ARTEX, an agentic AI penetration testing framework adopted by Chinese threat actors and flagged by the FBI/IC3 for enabling network breaches in under three hours. A backdoored ARTEX Docker image and exposed ML infrastructure credentials further expand the attack surface.

FBI NCIC Database Breach: Nationwide Police Intelligence Repository Compromised

The FBI is investigating an apparent breach of the National Crime Information Center (NCIC), a nationwide law enforcement intelligence repository managed by the CJIS Division. The compromise follows prior activity attributed to ShinyHunters targeting federal law enforcement infrastructure. Exposed data classes include active warrants, stolen property records, missing persons entries, and PII of suspects, victims, and personnel. Technical artifacts under review include NCIC query logs, CJIS network telemetry, authentication records, and ShinyHunters-associated IOCs. Impact includes degraded inter-agency intelligence sharing, officer safety risks, and potential follow-on attacks using stolen credentials.

Operation ShadowCloud: Microsoft SharePoint and GitHub Cloud-Native Backdoor Campaign

Operation ShadowCloud is an active cyber espionage campaign targeting government networks in Asia and Africa, identified by Kaspersky in early 2026. The campaign deploys two undocumented cloud-native backdoors, GitDoor and GraphDoor, which abuse GitHub and Microsoft SharePoint/Graph API for command-and-control (C2). By tunneling C2 traffic through legitimate SaaS infrastructure, the threat actor achieves persistent access and data exfiltration while blending with normal cloud operations, evading traditional network-based detection.

Microsoft agent-framework PyPI Supply Chain Compromise Steals Azure and CI Credentials

Microsoft’s official agent-framework Python package on PyPI was compromised on November 17, 2026, with malicious versions 1.26.1 and 1.26.2 published 33 minutes apart. The wheel-only releases embedded .pth persistence files (agent_framework_meta.pth, af_runtime_diag.pth) that executed a credential stealer on every Python interpreter startup, independent of package imports. The payload harvested Azure managed identity tokens, GitHub PATs, CI/CD secrets, environment variables, and dozens of credential files, exfiltrating compressed JSON to attacker-controlled webhook.site endpoints. Both versions remained live for approximately 55 minutes, poisoning container images, CI caches, and lockfiles built during the window.

Cairn AI Agent Framework Enables Autonomous Intrusions Against Semiconductor, Defense, and Civil Society Targets

The Cairn AI agent framework, an open-source LLM-driven penetration testing toolkit, was weaponized by a China-linked threat actor to conduct autonomous intrusions against semiconductor, defense, civil society, and higher education targets. Actors deployed Cairn agents to plan and execute exploit chains without human direction for nearly 24 hours, achieving persistence and server takeover on at least one university host. Amazon Security observed autonomous agent adaptation over this period, while Unit 42 and Aviatrix attributed the activity to a Chinese nexus and identified a suspected zero-day exploit against university server infrastructure supported by dedicated C2 infrastructure.

ARMORY CASTLE Exploits VMware vCenter Zero-Day CVE-2026-123890 for Admin Hijack and VM Data Theft

ARMORY CASTLE (UNC5221), a China-linked espionage APT, exploited CVE-2026-123890, a critical authentication bypass in the VMware vCenter Server directory service, to achieve administrator-level access. The CVSS 9.8 vulnerability allowed network-based compromise of the vCenter management console for ESXi hosts and virtual machines between early September and November 19, 2026. Actors hijacked admin accounts and exfiltrated files from a Windows VM. Broadcom disclosed the flaw and issued patches for all supported branches; no workaround exists.

Java4Shell: OpenJDK CVE-2026-114938 RCE Exploited via ARTEX and Chrome User-Agent Spoofing

Threat actors are actively exploiting CVE-2026-114938 (Java4Shell), an OpenJDK remote code execution vulnerability. Cloudflare Cloudforce One observed a two-stage pattern: initial reconnaissance from a source IP using a scanner-identifying User-Agent, followed approximately 17 hours later by exploit attempts spoofing Google Chrome User-Agent headers. The ARTEX framework obfuscates payloads using Base64 encoding to evade signature-based detection. GreyNoise, WatchTowr, and Cloudflare confirmed probing and exploitation. Cisco, Elastic, Red Hat, and Siemens products are impacted. Vendors including Cisco, Microsoft, and Wiz have issued advisories, patches, and Azure WAF mitigations.

Former Conti Affiliates Arrested in $40.5M International Vishing Campaign

Ukrainian Cyberpolice and the FBI arrested four suspected former Conti ransomware affiliates following a two-year investigation into a vishing operation that extorted $40.5 million from at least 28 organizations. Attackers used VoIP infrastructure, caller ID spoofing, and IT-support pretexting to harvest corporate credentials and gain VPN/RDP access. Post-compromise activity included lateral movement via Mimikatz, PsExec, and RDP, data exfiltration, and deployment of Conti-associated ransomware. Victims spanned oil, healthcare, and corporate sectors internationally. Extortion payments were processed through cryptocurrency wallets and dark-web negotiation channels.


LINK COPIED TO CLIPBOARD