CVE-2021-29602: TensorFlow DepthwiseConv 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 includes various operators, including the DepthwiseConv operator, which is used for depthwise separable convolution in neural networks. This operator is vulnerable to a division by zero error, which can be exploited by an attacker.

Vulnerability Detail

The vulnerability in the DepthwiseConv operator arises from a division by zero error in the implementation code. Specifically, the issue occurs when the fourth dimension of the input tensor is set to zero. An attacker can craft a malicious model that triggers this condition, potentially leading to a crash or other unintended behavior. The vulnerability affects versions of TensorFlow ranging from 0 to 2.4.2, 2.3.3, 2.2.3, and 2.1.4.

pip install --upgrade "tensorflow-cpu==2.21.0"

Workaround

As a workaround, users can ensure that the fourth dimension of the input tensor is always set to a non-zero value. This can be achieved by validating the input data and adjusting the dimensions if necessary. Here is an example code snippet that demonstrates the workaround:

import tensorflow as tf

# Validate input tensor dimensions
if input.shape[3] == 0:
    input = tf.reshape(input, (input.shape[0], input.shape[1], input.shape[2], 1))

# Continue with normal operations
...

Conclusion

The TensorFlow DepthwiseConv operator division by zero vulnerability can be mitigated by upgrading to the latest version of TensorFlow or applying the provided patches. It is crucial for users to promptly address this vulnerability to ensure the security and stability of their TensorFlow-based machine learning applications. For further support and assistance with vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow-cpu →

FAQ