CVE-2021-29587: TensorFlow GPU - SpaceToDepth TFLite Operator Division by Zero Vulnerability

Background

TensorFlow is an end-to-end open source platform for machine learning. It provides a comprehensive ecosystem of tools, libraries, and resources for developing and deploying machine learning models. TensorFlow GPU is a variant of TensorFlow that leverages GPU acceleration for faster computation. The SpaceToDepth TFLite operator is used for rearranging data in a tensor by spatially partitioning it into blocks and then concatenating them. This operator is susceptible to a division by zero vulnerability.

Vulnerability Detail

The vulnerability (CVE-2021-29587) arises from the lack of a check for zero before division in the Prepare step of the SpaceToDepth TFLite operator in TensorFlow. An attacker can exploit this vulnerability by crafting a malicious model that sets the params->block_size parameter to zero. When the model is executed, the division by zero occurs, leading to potential crashes, unexpected behavior, or even remote code execution. The vulnerability affects TensorFlow versions 2.1.4 to 2.4.3, as well as TensorFlow GPU versions 0 to 2.4.3.

pip install --upgrade "tensorflow-gpu==2.12.0"

Workaround

As a temporary workaround, users can manually modify the affected code in the TensorFlow source files. Specifically, the space_to_depth.cc file located at tensorflow/lite/kernels/space_to_depth.cc needs to be edited. The Prepare function should be modified to include a check for zero before performing the division operation. Here is an example of the modified code:

double block_size = params->block_size;
if (block_size == 0) {
  // Handle the division by zero case
  // Perform necessary actions
} else {
  // Perform the division operation
  // Continue with the normal execution
}

Conclusion

The division by zero vulnerability in the SpaceToDepth TFLite operator of TensorFlow GPU can be exploited by an attacker to cause crashes, unexpected behavior, or even remote code execution. Users are advised to update their TensorFlow installations to the latest patched version and follow the official documentation for guidance. For further support on vulnerability remediation, contact DevNack.com.

See every vulnerability recorded for tensorflow-gpu →

FAQ