Executive brief
Feast is an open-source feature store used to manage and serve machine learning data. A security flaw in its Kubernetes component allows an attacker who can modify certain configuration files to execute malicious commands on the system. This could lead to a complete takeover of the data processing environment, potentially resulting in data theft or the corruption of machine learning models.
Technical details
A remote code execution vulnerability exists in Feast due to the use of the unsafe `yaml.load()` function with `yaml.Loader` in the Kubernetes materializer job. Specifically, the component `feast/sdk/python/feast/infra/compute_engines/kubernetes/main.py` insecurely deserializes `/var/feast/feature_store.yaml` and `/var/feast/materialization_config.yaml`. An attacker with local access or the ability to modify these YAML files can instantiate arbitrary Python objects to execute OS commands on the worker pod. This can occur before configuration validation, potentially leading to cluster takeover or supply-chain sabotage. The issue is resolved in version 0.54.0 by switching to `yaml.safe_load()`.
Affected products
- feast-dev Feast < 0.54.0
- Red Hat OpenShift AI 2.25 < 2.25.5
Timeline
- 2026-01-01: disclosed: Vulnerability reported via huntr.dev
- 2026-01-01: advisory: CVE-2025-11157 published
- 2026-04-23: patched: Red Hat released fix for OpenShift AI 2.25.5
References
- https://github.com/feast-dev/feast/commit/b2e37ff37953b68ae833f6874ab5bc510a4ca5fb
- https://huntr.com/bounties/46d4d585-b968-4a76-80ce-872bc5525564
- https://access.redhat.com/errata/RHSA-2026:10184
- https://access.redhat.com/security/cve/CVE-2025-11157
- https://bugzilla.redhat.com/show_bug.cgi?id=2426574
- https://security.access.redhat.com/data/csaf/v2/vex/2025/cve-2025-11157.json