CVE-2021-29573: TensorFlow GPU - Division by 0 in MaxPoolGradWithArgmax

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 vulnerability affects the implementation of tf.raw_ops.MaxPoolGradWithArgmax in TensorFlow GPU.

Vulnerability Detail

The vulnerability (CVE-2021-29573) arises from a division by 0 in the implementation of tf.raw_ops.MaxPoolGradWithArgmax. The affected code fails to validate that the batch dimension of the tensor is non-zero before dividing by this quantity. As a result, when the batch dimension is 0, a division by 0 occurs, leading to a potential crash or undefined behavior. The vulnerability can be exploited by an attacker to cause a denial-of-service (DoS) condition or potentially execute arbitrary code.

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

Workaround

As a temporary workaround, users can modify the code to explicitly check if the batch dimension is non-zero before calling tf.raw_ops.MaxPoolGradWithArgmax. Here is an example of how the code can be modified:

import tensorflow as tf

# Check if batch dimension is non-zero
if tensor.shape[0] != 0:
    result = tf.raw_ops.MaxPoolGradWithArgmax(...)
else:
    # Handle the case when batch dimension is 0
    result = ...

Conclusion

The vulnerability (CVE-2021-29573) in TensorFlow GPU's tf.raw_ops.MaxPoolGradWithArgmax function can result in a division by 0 error, potentially leading to a crash or undefined behavior. It is crucial for users to update their TensorFlow GPU installations to the patched versions once they are released. Regularly checking for updates and applying them promptly is recommended to ensure the security and stability of the TensorFlow GPU environment. For further support on vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow-gpu →

FAQ