Impact
Certain TFLite models that were created using TFLite model converter would crash when loaded in the TFLite interpreter. The culprit is that during quantization the scale of values could be greater than 1 but code was always assuming sub-unit scaling.
Thus, since code was calling QuantizeMultiplierSmallerThanOneExp, the TFLITE_CHECK_LT assertion would trigger and abort the process.
Patches
We have patched the issue in GitHub commit a989426ee1346693cc015792f11d715f6944f2b8.
The fix will be included in TensorFlow 2.9.0. We will also cherrypick this commit on TensorFlow 2.8.1, TensorFlow 2.7.2, and TensorFlow 2.6.4, as these are also affected and still in supported range.
For more information
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
Attribution
This vulnerability has been reported externally via a GitHub issue.
References
- https://github.com/tensorflow/tensorflow/security/advisories/GHSA-8wwm-6264-x792
- https://nvd.nist.gov/vuln/detail/CVE-2022-29212
- https://github.com/tensorflow/tensorflow/issues/43661
- https://github.com/tensorflow/tensorflow/commit/a989426ee1346693cc015792f11d715f6944f2b8
- https://github.com/tensorflow/tensorflow/blob/f3b9bf4c3c0597563b289c0512e98d4ce81f886e/tensorflow/lite/kernels/internal/quantization_util.cc#L114-L123
- https://github.com/tensorflow/tensorflow/releases/tag/v2.6.4
- https://github.com/tensorflow/tensorflow/releases/tag/v2.7.2
- https://github.com/tensorflow/tensorflow/releases/tag/v2.8.1
- https://github.com/tensorflow/tensorflow/releases/tag/v2.9.0
- https://github.com/advisories/GHSA-8wwm-6264-x792