supply chain attack

Bilingual

Shai-Hulud and the risks of external dependencies
2 months ago
- Shai-Hulud is an npm supply-chain malware by TeamPCP, named after the sandworms from Dune, which compromised 500 packages, breached hundreds of organizations, stole 300 GB of data, and exfiltrated ~500,000 credentials.
- The malware's source code was released on May 11; analysis shows it uses simple tricks like credential theft and string obfuscation rather than advanced exploits, with inconsistent, buggy, and incomplete code.
- It works by injecting malicious dependencies or preinstall scripts via OIDC or CI pipelines, stealing credentials (AWS, SSH, GitHub tokens, etc.), and exfiltrating data through encrypted channels or compromised GitHub repos.
- Key weaknesses in modern development exploited: blind trust in dependencies and ease of importing code; risks include editor integrations and CI systems running malicious scripts.
- Recommendations include reducing dependencies, using sandboxed runtimes like Deno, enforcing zero-trust permissions, and adopting stable release channels for auditing; npm now blocks install scripts from dependencies to mitigate such attacks.
Telnyx, LiteLLM and Axios: the supply chain crisis
2 months ago
- A wave of supply chain attacks targeted popular npm packages (Telnyx, LiteLLM, axios) by injecting trojans to steal sensitive data, exploiting stolen credentials from previous attacks.
- Attackers used compromised developer credentials to publish malicious versions, creating a vicious cycle of infections that amplified the reach of the attack.
- LLMs are suspected to be accelerants, making it easier to find vulnerabilities and build sophisticated exploits, such as hiding code in audio files.
- Current mitigations (short-lived tokens, package delays, staged publishing) are insufficient; they introduce delays or can be bypassed if developer machines are compromised.
- The author proposes shifting desktop and server OSes to a sandboxed model similar to iOS/Android, where each app has limited permissions to reduce the blast impact of attacks.
- Governments should allocate resources to help open source projects adopt sandboxing and other security measures to address the growing threat.
CPU-Z Backdoor: Detecting Malicious Implants at Scale
2 months ago
- AI writing code introduces security risks as code review becomes impossible due to speed.
- Software supply chain attacks are prevalent; attackers exploit vulnerabilities within days of disclosure.
- RevEng.AI analyzes binary artifacts to detect malicious components before production deployment.
- Case study: CPU-Z supply chain attack distributed trojanized installers containing STX RAT malware.
- The multi-stage infection chain used DLL loaders, API resolution, shellcode injection, and final payload delivery.
- RevEng.AI's Reverser platform automates triage, decompilation, and IoC extraction for faster analysis.
- The analysis mapped capabilities to MITRE ATT&CK and generated a YARA rule for automated remediation.
Exploiting LLM Agent Supply Chains via Payload-Less Skills
3 months ago
- Introduces Semantic Compliance Hijacking (SCH), a payload-less supply chain attack that uses natural language instructions to make LLM agents generate and execute unauthorized code.
- Current security scanning tools fail to detect SCH because it lacks explicit code payloads and harmful AST signatures, achieving a 0.00% detection rate.
- The attack was tested across three agent frameworks and three foundation models, with peak success rates of 77.67% for confidentiality breaches and 67.33% for RCE.
- Multi-Skill Automated Optimization (MS-AO) further enhances the attack's effectiveness.
- Highlights the need to move from signature-based detection to semantic intent validation in LLM agent supply chains.

supply chain attack

Shai-Hulud and the risks of external dependencies

Telnyx, LiteLLM and Axios: the supply chain crisis

CPU-Z Backdoor: Detecting Malicious Implants at Scale

Exploiting LLM Agent Supply Chains via Payload-Less Skills