CVE-2021-29574: TensorFlow GPU - Null Pointer Dereference in MaxPool3DGradGrad

Sponsored Habilume Habit tracker with a companion that grows with you Complete habits, earn Lume Points, and watch your Lume grow as you stay consistent. App Store Google Play

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 the power of GPUs for accelerated computation. It allows users to train and run machine learning models on GPU hardware, enabling faster and more efficient processing of large datasets. However, a vulnerability has been identified in the TensorFlow GPU package that can lead to a null pointer dereference in the MaxPool3DGradGrad implementation.

Vulnerability Detail

The vulnerability (CVE-2021-29574) arises from the improper handling of empty tensors in the tf.raw_ops.MaxPool3DGradGrad function of TensorFlow GPU. The implementation fails to validate whether the three tensor inputs are empty before accessing their elements. As a result, if any of the tensors is empty, the code attempts to dereference a null pointer, leading to undefined behavior. An attacker can exploit this vulnerability by supplying empty tensors as inputs, potentially causing a crash or other unintended consequences. The issue has been identified and fixed in TensorFlow 2.5.0, and patches have been backported to earlier versions.

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

Workaround

If updating or patching the TensorFlow GPU package is not immediately feasible, a temporary workaround is to validate the tensor inputs manually before calling the MaxPool3DGradGrad function. By checking if the tensors are empty, you can avoid triggering the null pointer dereference. Here is an example of how to perform the validation in Python:

def max_pool_3d_grad_grad_workaround(tensor1, tensor2, tensor3):
    if tensor1.size == 0 or tensor2.size == 0 or tensor3.size == 0:
        # Handle empty tensors
        return None
    else:
        # Call the MaxPool3DGradGrad function
        return tf.raw_ops.MaxPool3DGradGrad(tensor1, tensor2, tensor3)

Conclusion

In conclusion, the TensorFlow GPU package is vulnerable to a null pointer dereference issue in the MaxPool3DGradGrad implementation. This vulnerability can be exploited by providing empty tensors as inputs, potentially leading to crashes or other unintended consequences. It is crucial to update the package to version 2.5.0 or apply the provided patches to mitigate the vulnerability. For further support on vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow-gpu →

FAQ