Unknown cybercriminals have successfully infiltrated two legitimate MemTensor packages within the npm and Python Package Index (PyPI) repositories. This breach leverages a specific Go-based malware named ‘sckit’, which is designed to operate across Windows, Linux, and macOS environments.
Details of the Compromised Packages
Reports from security firms such as Aikido, SafeDep, Socket, and StepSecurity highlight that the npm package versions 0.1.21, 0.1.23, and 0.1.25 are affected. These versions integrate a concealed Go payload into an AI memory module, activating the malicious code during memory recall events or when gateway agents initialize.
In the PyPI ecosystem, the threat emerges when the ‘memos’ module is used in an application, triggering the associated Go binary. The ultimate aim of these attacks is to deploy a cross-platform credential-stealing payload, targeting sensitive data from cloud services, source-code platforms, and developer tools, subsequently transmitting this data to an external server.
Targets and Impact
Socket’s analysis indicates that potential targets include npm, PyPI, GitHub, GitLab, AWS, and various credential files and environment variables containing sensitive information like tokens and API keys. The malware is capable of extracting AWS keys, GitHub and GitLab tokens, and other significant credentials.
SafeDep’s investigation into the supply chain attack reveals that the attackers exploited MemTensor’s GitHub Actions release pipelines to obtain publish tokens, facilitating unauthorized npm or PyPI token access. This breach allows the malware to proliferate like a worm across GitHub and other package repositories.
Security Recommendations and Outlook
It is crucial for developers and organizations using these packages to enforce immediate security measures. This includes reverting to baseline package versions—0.1.20 for npm and 2.0.33 for PyPI—rotating exposed secrets, terminating any ‘sckit’ processes, and blocking the ‘skyleen[.]fr’ domain.
As the MemOS Cloud plugin connects to various runtime environments to process and store memory data, it inadvertently becomes a vehicle for credential theft. This is particularly concerning on developer machines and in automated environments where sensitive credentials are routinely handled.
The situation underscores the need for vigilant supply chain security practices and highlights the ongoing risks posed by sophisticated cyber threats targeting software development ecosystems.
