| Package | Affected versions | Patched versions |
|---|---|---|
| tensorflow | ≥ 2.6.0, < 2.6.3≥ 2.7.0, < 2.7.1< 2.5.3 | 2.6.32.7.12.5.3 |
| tensorflow-cpu | ≥ 2.6.0, < 2.6.3≥ 2.7.0, < 2.7.1< 2.5.3 | 2.6.32.7.12.5.3 |
| tensorflow-gpu | ≥ 2.6.0, < 2.6.3≥ 2.7.0, < 2.7.1< 2.5.3 | 2.6.32.7.12.5.3 |
| tensorflow | = 2.7.0≥ 2.6.0, ≤ 2.6.2≤ 2.5.2 |
An attacker can craft a TFLite model that would trigger a division by zero in BiasAndClamp implementation:
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.
We have patched the issue in GitHub 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.
Please consult our security guide for more information regarding the security model and how to contact us with issues and questions.
This vulnerability has been reported by Wang Xuan of Qihoo 360 AIVul Team.
Running one of these products? Scan your site to see whether it exposes a vulnerable version.
Scan my site free