Core dump when loading TFLite models with quantization in TensorFlow

Summary

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.

Timeline

1. May 20, 2022, 11:15 PM UTCCVE published
2. Apr 22, 2025, 3:46 PM UTCCISA SSVC assessment

   Exploitation: poc · Automatable: no · Technical impact: partial

3. Apr 22, 2025, 5:57 PM UTCRecord last updated

Affected versions

• tensorflow tensorflow

  fixed in 2.6.4; From 2.7.0rc0 · fixed in 2.7.2; From 2.8.0rc0 · fixed in 2.8.1; From 2.9.0rc0 · fixed in 2.9.0

• tensorflowPyPI

  fixed in 2.6.4

• tensorflowPyPI

  From 2.7.0 · fixed in 2.7.2

• tensorflowPyPI

  From 2.8.0 · fixed in 2.8.1

• tensorflow-cpuPyPI

  fixed in 2.6.4

• tensorflow-cpuPyPI

  From 2.7.0 · fixed in 2.7.2

• tensorflow-cpuPyPI

  From 2.8.0 · fixed in 2.8.1

• tensorflow-gpuPyPI

  fixed in 2.6.4

• tensorflow-gpuPyPI

  From 2.7.0 · fixed in 2.7.2

• tensorflow-gpuPyPI

  From 2.8.0 · fixed in 2.8.1

• tensorflowPyPI

  fixed in 2.6.4; From 2.7.0 · fixed in 2.7.2; From 2.8.0 · fixed in 2.8.1

• tensorflow-cpuPyPI

  fixed in 2.6.4; From 2.7.0 · fixed in 2.7.2; From 2.8.0 · fixed in 2.8.1

• tensorflow-gpuPyPI

  fixed in 2.6.4; From 2.7.0 · fixed in 2.7.2; From 2.8.0 · fixed in 2.8.1

References 8