PyPI-hosted Microsoft Agent Framework Library Hijacked to Deliver Credential‑Theft Malware
Researchers in security have discovered that the genuine Microsoft Agent Framework package on the Python Package Index (PyPI) had been modified to embed malicious code that can siphon user credentials and cloud service tokens.
The compromised package—widely used by developers to embed Microsoft’s AI agents—was identified as carrying a payload that covertly harvested authentication information for several developer‑focused services such as Azure, Amazon Web Services, GitHub, OpenAI and Docker.
Upon installation of the tainted version, the concealed component ran at runtime, pulling stored tokens and sending them to a server under the attacker’s control. This exfiltrated data could then be leveraged to obtain unauthorized entry into cloud resources, code repositories, and other services linked to the victim’s accounts.
Microsoft together with PyPI maintainers acted swiftly once the breach was reported, pulling the malicious releases and reinstating the authentic package. Developers who downloaded the compromised version are urged to rotate any impacted credentials and inspect access logs for any anomalous activity.
The episode joins an expanding tally of supply‑chain assaults on open‑source ecosystems, where threat actors hijack popular libraries to affect many users with little effort. PyPI, similar to other package registries, operates on a trust model that presumes published packages are genuine, rendering it a tempting conduit for attackers aiming to penetrate development pipelines.
Specialists advise developers to confirm package integrity via hash verification, use virtual environments, and keep a close eye on dependency updates. Continuous vigilance and swift incident response remain essential as the software supply chain stays a prime target for cyber threats.
Comments (0)
Be the first to comment.
Join the discussion