Skip to content
AI Sec Watch
MediumVulnerability

CVE-2020-15265: Tensorflow out-of-bounds access via invalid axis in quantize_and_dequantize

Source
NVD(opens in a new tab)
Identifier
CVE-2020-15265
Published
Record updated
View JSON
Affected
  • tensorflow < 2.4.0
  • tensorflow-cpu < 2.4.0
  • tensorflow-gpu < 2.4.0
Fixed in
2.4.0
  • tensorflow 2.4.0 released December 14, 2020, 53 days after disclosure
  • tensorflow-cpu 2.4.0 released December 14, 2020, 53 days after disclosure
  • tensorflow-gpu 2.4.0 released December 14, 2020, 53 days after disclosure
Known exploitation
Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
EPSS
0.9%

Summary

CVE-2020-15265 affects TensorFlow versions before 2.4.0. Passing an invalid `axis` value to `tf.quantization.quantize_and_dequantize` makes the C++ kernel access a dimension outside the input tensor's rank. The bounds check is only a DCHECK, which is a no-op in normal builds, so the out-of-bounds array access can cause a segfault.

Mitigation

The issue is patched in commit eccb7ec454e6617738554a255d77f08e60ee0808, and TensorFlow 2.4.0 will be released containing the patch. TensorFlow nightly packages built after this commit also have the issue resolved.

Affected packages in the Exposure Registry

Matched by package name and ecosystem. Each entry shows whether the package delegates to a language model and how many tracked packages depend on it.

Related items