Core dump when loading TFLite models with quantization in TensorFlow
tensorflow
Affected versions
| Product | Affected | Fixed in |
|---|---|---|
| tensorflow PyPI |
|
|
| tensorflow-cpu PyPI |
|
tensorflow
| Product | Affected | Fixed in |
|---|---|---|
| tensorflow PyPI |
|
|
| tensorflow-cpu PyPI |
|
≥ 2.7.0 · < 2.7.2≥ 2.8.0 · < 2.8.1
|
| tensorflow-gpu PyPI |
|
|
TensorFlow is an open source platform for machine learning. Prior to versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4, 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. Versions 2.9.0, 2.8.1, 2.7.2, and 2.6.4 contain a patch for this issue.
From the CVE record · GitHub_M
Certain models created with the TFLite model converter can trigger an assertion failure when loaded in the TFLite interpreter. An attacker who can get an affected application to load such a model could cause its process to abort; the input does not specify a network path or authentication requirement.
Engineers using PyPI tensorflow, tensorflow-cpu, or tensorflow-gpu versions below 2.6.4, from 2.7.0 to below 2.7.2, or from 2.8.0 to below 2.8.1 should check their deployments.