First published: Fri Feb 04 2022(Updated: )
### Impact An attacker can craft a TFLite model that would trigger a division by zero in [`BiasAndClamp` implementation](https://github.com/tensorflow/tensorflow/blob/5100e359aef5c8021f2e71c7b986420b85ce7b3d/tensorflow/lite/kernels/internal/common.h#L75): ```cc inline void BiasAndClamp(float clamp_min, float clamp_max, int bias_size, const float* bias_data, int array_size, float* array_data) { // ... TFLITE_DCHECK_EQ((array_size % bias_size), 0); // ... } ``` There is no check that the `bias_size` is non zero. ### Patches We have patched the issue in GitHub commit [8c6f391a2282684a25cbfec7687bd5d35261a209](https://github.com/tensorflow/tensorflow/commit/8c6f391a2282684a25cbfec7687bd5d35261a209). The fix will be included in TensorFlow 2.8.0. We will also cherrypick this commit on TensorFlow 2.7.1, TensorFlow 2.6.3, and TensorFlow 2.5.3, as these are also affected and still in supported range. ### For more information Please consult [our security guide](https://github.com/tensorflow/tensorflow/blob/master/SECURITY.md) for more information regarding the security model and how to contact us with issues and questions. ### Attribution This vulnerability has been reported by Wang Xuan of Qihoo 360 AIVul Team.
Credit: security-advisories@github.com security-advisories@github.com
Affected Software | Affected Version | How to fix |
---|---|---|
Google TensorFlow | <=2.5.2 | |
Google TensorFlow | >=2.6.0<=2.6.2 | |
Google TensorFlow | =2.7.0 | |
pip/tensorflow-gpu | =2.7.0 | 2.7.1 |
pip/tensorflow-gpu | >=2.6.0<2.6.3 | 2.6.3 |
pip/tensorflow-gpu | <2.5.3 | 2.5.3 |
pip/tensorflow-cpu | =2.7.0 | 2.7.1 |
pip/tensorflow-cpu | >=2.6.0<2.6.3 | 2.6.3 |
pip/tensorflow-cpu | <2.5.3 | 2.5.3 |
pip/tensorflow | =2.7.0 | 2.7.1 |
pip/tensorflow | >=2.6.0<2.6.3 | 2.6.3 |
pip/tensorflow | <2.5.3 | 2.5.3 |
Sign up to SecAlerts for real-time vulnerability data matched to your software, aggregated from hundreds of sources.