CCTR.4.JAN.26
Monday morning cyber coffee read CCTR.4.JAN.26.
Check Point Research has identified a previously unseen Linux-based malware framework, VoidLink, designed to provide long-term, stealthy access to cloud and container environments.
It incorporates advanced built-in evasion capabilities and appears to be developed by Chinese-affiliated actors.
Based on my experience running purple team engagements, EDR coverage is often strong in traditional corporate environments but materially weaker across Linux and container platforms.
Reduced telemetry and limited behavioural detection make these environments attractive operating space for advanced threat actors seeking stealth and long-term access.
Improve detection coverage for Linux including suspicious inter-container communication.
Monitor cloud control planes for unusual API usage, credential abuse and lateral movement across workloads and accounts.
Validate assumptions through purple teaming focused on Linux, container and cloud attack paths, rather than testing your corporate network centric controls.
https://research.checkpoint.com/2026/voidlink-the-cloud-native-malware-framework/
Three high-severity vulnerabilities have been identified in Anthropic’s official MCP Git server, allowing attackers to read or delete arbitrary files and achieve remote code execution via prompt injection. By influencing what an AI assistant reads, attackers can chain these flaws to break filesystem boundaries and execute commands.
The issues have been patched but the findings highlight a broader risk where AI tooling can turn untrusted content into a direct execution path.
If security boundaries fail in reference MCP implementations, organisations should assume similar risks exist across their AI and agentic tooling stack.
Review MCP and agent permissions, especially filesystem and command execution capabilities.
Validate AI-assisted workflows through purple teaming, focusing on prompt injection and cross-tool abuse scenarios.
https://thehackernews.com/2026/01/three-flaws-in-anthropic-mcp-git-server.html
Ransomware attackers targeting a Fortune 100 financial services organisation have been observed using a new Windows malware strain, PDFSider, which abuses signed binaries, memory-only execution and encrypted DNS-based C2 to significantly reduce visibility for traditional endpoint controls. Although linked to Qilin ransomware activity, the malware employs tradecraft more commonly associated with APTs.
Malware like this blurs the line between ransomware and espionage tradecraft and highlights that there are strong operational reasons for ransomware actors to adopt these techniques.
Organisations should test detection beyond traditional malware signatures and assume adversaries are already operating below the noise floor.
Last updated