Track East Asia cyber, AI, cloud, and infrastructure risk before it becomes an incident.
The Friendly Fire proof-of-concept demonstrates how attackers can weaponize AI coding agents through prompt injection in project documentation, achieving remote code execution by exploiting the agent's inability to distinguish between data and executable instructions without modifying the agent itself. Read more
ASEC’s analysis of a 2026 phishing email campaign by the Larva-24009 threat actor details how LNK files disguised as legitimate documents deploy PowerShell backdoors, QuasarRAT, UltraVNC, and NirSoft tools for credential theft, with persistence via scheduled tasks and exfiltration through the Telegram API, targeting users in Korea and globally. Read more
While the state-sponsored group deployed backdoors (Struggle/SIGNBT 3.0 and Brandoor/COPPERHEDGE) for espionage, Gunra ransomware was used in parallel attacks for data encryption and exfiltration. Overlapping indicators—including SSH key fingerprints, network infrastructure, and watering hole domains—suggest shared TTPs, though ASEC concludes no definitive collaboration has been proven. The campaign, named 'Operation Double Barrel,' highlights the risk of dual-use exploits in critical financial software supply chains. Read more
CISA has added CVE-2025-68686 (Fortinet FortiOS information exposure) and CVE-2026-16812 (Arista VeloCloud Orchestrator command injection) to its Known Exploited Vulnerabilities Catalog based on evidence of active exploitation, reinforcing BOD 26-04 requirements for federal agencies and urging all organizations to prioritize patching. Read more
OpenAI has released the command-line interface and TypeScript SDK for Codex Security as open-source software, enabling developers to scan local code and Git changes for vulnerabilities and integrate checks into pre-commit and CI workflows, though actual analysis relies on OpenAI cloud services and requires authentication via ChatGPT login or API key. Read more
A vulnerability in ABB's KNX Update Tool allows attackers with physical bus access to compromise legacy KNX devices lacking integrity checks, rendering them unusable or altering behavior, with no software fix possible due to outdated protocol design. Read more
A critical vulnerability (CVE-2026-7891) in Siemens Mendix Runtime allows privilege escalation and unauthorized access to sensitive user data due to inadequate documentation of System.User entity behavior, enabling misconfigured access rules that expose all records to anonymous users. Read more
A new large language model framework integrates hierarchical, tree-based, and tokenization-based retrieval to improve accuracy and traceability in China's Multi-Level Protection Scheme (MLPS) compliance analysis, addressing limitations of general-purpose LLMs in standards-intensive cybersecurity governance. Read more
AhnLab ASEC details a four-stage in-memory loader chain for AtlasRAT, a Windows RAT using Delphi-based Flash Player lures, TLS-ChaCha20 C2, offline keylogging, and WeChat.exe DLL injection, with evidence pointing to a builder-based framework rather than a single operator, highlighting implications for regional threat monitoring. Read more
A Chinese national AI agent identity system launching in Q3 2026 uses split-knowledge binding to link agents to verified legal principals without revealing identities to business participants, requiring dual-agency compulsion for re-identification by legal authorities. Read more