FPE in LSH in TFLite

Overview

Impact

An attacker can craft a TFLite model that would trigger a division by zero error in LSH implementation.

int RunningSignBit(const TfLiteTensor* input, const TfLiteTensor* weight,
                   float seed) {
  int input_item_bytes = input->bytes / SizeOfDimension(input, 0);
  // ...
}

There is no check that the first dimension of the input is non zero.

Patches

We have patched the issue in GitHub commit 0575b640091680cfb70f4dd93e70658de43b94f9.

The fix will be included in TensorFlow 2.6.0. We will also cherrypick thiscommit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, 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 by Yakun Zhang of Baidu Security.

Affected packages

• tensorflow < 2.3.4
• tensorflow >= 2.4.0, < 2.4.3
• tensorflow >= 2.5.0, < 2.5.1
• tensorflow-cpu < 2.3.4
• tensorflow-cpu >= 2.4.0, < 2.4.3
• tensorflow-cpu >= 2.5.0, < 2.5.1
• tensorflow-gpu < 2.3.4
• tensorflow-gpu >= 2.4.0, < 2.4.3
• tensorflow-gpu >= 2.5.0, < 2.5.1

Remediation

Upgrade to a patched release:

• tensorflow 2.3.4
• tensorflow 2.4.3
• tensorflow 2.5.1
• tensorflow-cpu 2.3.4
• tensorflow-cpu 2.4.3
• tensorflow-cpu 2.5.1
• tensorflow-gpu 2.3.4
• tensorflow-gpu 2.4.3
• tensorflow-gpu 2.5.1