PYSEC-2026-1026: Missing validation in `QuantizedConv2D` results in undefined behavior in TensorFlow
TensorFlow is an open source platform for machine learning. Prior to versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4, the implementation of tf.rawops.QuantizedConv2D does not fully validate the input arguments. In this case, references get bound to nullptr for each argument that is empty. Versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4 contain a patch for this issue.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
TensorFlowto a version that resolves this vulnerability.Fixed in 2.9.0
Event History
Frequently Asked Questions
Which TensorFlow versions are affected?
TensorFlow versions before 2.6.4, 2.7.2, 2.8.1, and 2.9.0 are affected. The issue is patched in those respective releases.
What does an attacker need to trigger the issue?
The attacker needs local access and low privileges. Triggering the issue involves calling tf.raw_ops.QuantizedConv2D with empty input arguments, which can cause references to bind to nullptr.
What is the expected security impact?
The reported impact is availability only: the vulnerability is associated with undefined behavior and null pointer dereference. No confidentiality or integrity impact is indicated by the provided severity vector.
What should be done if the deployment is vulnerable?
Upgrade TensorFlow to a patched release: 2.6.4, 2.7.2, 2.8.1, 2.9.0, or a later applicable version. Until then, avoid allowing untrusted or insufficiently validated calls to QuantizedConv2D with empty arguments.