Microsoft Qlib users should review their installed version and the files their research workflows load following a November 2 advisory from the Zero Day Initiative. The notice identifies a code execution vulnerability in pickle-data handling and lists version 0.9.8 as the fix.
Qlib is Microsoft’s open-source platform for quantitative investment research. Its project documentation describes workflows spanning data processing, machine-learning model training and backtesting. That makes the origin of saved research artifacts relevant when teams exchange or reload them.
A malicious file still needs a path into the workflow
ZDI attributes the flaw to insufficient validation during deserialization and says successful exploitation runs code with the current user’s privileges. The target must interact with malicious content, such as opening a file or visiting a page.
The advisory assigns a CVSS score of 7.8, with local attack-vector and required-user-interaction metrics. It does not describe an internet-facing Qlib service that can be compromised simply by sending it an unauthenticated network request.
Check compatibility before restoring saved artifacts
The vendor’s 0.9.8 release notes, dated October 8, describe broader security changes, including restrictions on executable recorder artifacts. Supported data-only reads and fresh training remain available without an opt-in; restoring executable objects from verified storage requires an explicit trust decision.
Microsoft warns that existing workflows may need configuration changes and directs users to its migration guides before production upgrades. Verify who produced an artifact and who can replace it before enabling trusted loading. The release supplies source archives; it tells users to check PyPI separately for package availability.
Sources: ZDI-26-849 advisory; Qlib 0.9.8 release notes; Microsoft Qlib project documentation.
Image: AI-generated editorial illustration created for DaySec, not a photograph of the incident.
