Phying News
P非影 中文
Curated security research, vulnerabilities, advisories and tools for practitioners.

LiteLLM Supply Chain Compromise

netspi.com | incident | #ai-security | #supply-chain | #incident-response | #credential-theft | #pypi | #llm | #cicd | #litellm

Summary

Malicious PyPI builds of LiteLLM 1.82.7/1.82.8, published via a compromised maintainer account, exfiltrated API keys and system data, exposing a user's entire LLM provider surface.
Published
Collected

original ↗

Related coverage

incident · hunt.io
33K Exposed LiteLLM Deployments and the C2 Servers Behind TeamPCP's Supply Chain Attack
After TeamPCP trojanized LiteLLM's PyPI packages, Hunt.io counted 33,688 exposed deployments and traced dual C2—AdaptixC2 and Havoc—plus a third server tied to the exfiltration domain.
blog · wiz.io
Three’s a Crowd: TeamPCP trojanizes LiteLLM in Continuation of Campaign
TeamPCP trojanized LiteLLM with malicious PyPI versions 1.82.7 and 1.82.8: the payloads abuse Python's .pth mechanism to run base64 payloads at startup, exfiltrating cloud and CI/CD credentials.
blog · netspi.com
3 Software Supply Chain Risks in 2023
Software supply chain attacks in 2023 cluster into three categories: package manager attacks like dependency confusion and typosquatting, developer compromise, and breaches at large software vendors.
blog · projectblack.io
Bypassing EDR with Local AI
A pentest shows Claude refuses to build an LSASS dumper, but local DeepSeek and uncensored Qwen models produce a working XOR-encrypted minidumper, making AI-assisted EDR evasion nearly effortless.
blog · wiz.io
The Infostealer Incursion: How Stolen Credentials Breach Cloud, Code, and AI Environments
Wiz Research analyzes NordStellar logs to map credentials harvested by infostealers like Lumma and RedLine, showing how stolen keys and tokens unlock AWS, Azure, GitHub, and AI platforms.
blog · specterops.io
AI for Offensive Security: What Works, What Does Not, and How to Adopt It
SpecterOps maps where AI helps offensive security (recon, tooling, vuln discovery) and where it still fails, stressing that results hinge on the harness: context, tools, state, independent validation, and human judgment.
back