CVE-2021-37682: Use of unitialized value in TFLite
(updated )
All TFLite operations that use quantization can be made to use unitialized values. For example:
const auto* affine_quantization =
reinterpret_cast<TfLiteAffineQuantization*>(
filter->quantization.params);
The issue stems from the fact that quantization.params
is only valid if quantization.type
is different that kTfLiteNoQuantization
. However, these checks are missing in large parts of the code.
References
- github.com/advisories/GHSA-4c4g-crqm-xrxw
- github.com/tensorflow/tensorflow
- github.com/tensorflow/tensorflow/blob/460e000de3a83278fb00b61a16d161b1964f15f4/tensorflow/lite/kernels/depthwise_conv.cc
- github.com/tensorflow/tensorflow/commit/4a91f2069f7145aab6ba2d8cfe41be8a110c18a5
- github.com/tensorflow/tensorflow/commit/537bc7c723439b9194a358f64d871dd326c18887
- github.com/tensorflow/tensorflow/commit/8933b8a21280696ab119b63263babdb54c298538
- github.com/tensorflow/tensorflow/security/advisories/GHSA-4c4g-crqm-xrxw
- nvd.nist.gov/vuln/detail/CVE-2021-37682
Detect and mitigate CVE-2021-37682 with GitLab Dependency Scanning
Secure your software supply chain by verifying that all open source dependencies used in your projects contain no disclosed vulnerabilities. Learn more about Dependency Scanning →