CVE-2021-29566: TensorFlow-GPU Dilation2DBackpropInput Heap Overflow

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 training and inference.

Vulnerability Detail

The vulnerability (CVE-2021-29566) exists in the tf.raw_ops.Dilation2DBackpropInput function of TensorFlow-GPU. By passing invalid arguments to this function, an attacker can trigger a heap overflow. The implementation of the function does not perform proper bounds checking before writing to the output array, allowing the attacker to write outside the bounds of heap allocated arrays. While the loop indices h_out and w_out are guaranteed to be within range, there are no similar guarantees for h_in_max and w_in_max. This vulnerability can be exploited to execute arbitrary code or cause a denial-of-service condition.

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

Workaround

As a temporary workaround, users can avoid passing invalid arguments to the tf.raw_ops.Dilation2DBackpropInput function. Ensure that the input arguments are within the expected range to prevent triggering the heap overflow. Here is an example of how to validate the input arguments:

import tensorflow as tf

def safe_dilation2d_backprop_input(input_shape, filter_shape, out_backprop, strides, padding):
    h_in_max, w_in_max = input_shape
    h_filter, w_filter = filter_shape
    h_out, w_out = tf.raw_ops.Dilation2DBackpropInput(input_sizes=[h_in_max, w_in_max], filter_sizes=[h_filter, w_filter], out_backprop=out_backprop, strides=strides, padding=padding)
    if h_out > h_in_max or w_out > w_in_max:
        raise ValueError('Invalid input arguments')
    return h_out, w_out

Conclusion

The vulnerability (CVE-2021-29566) in TensorFlow-GPU's tf.raw_ops.Dilation2DBackpropInput function allows an attacker to trigger a heap overflow. It is crucial to update to the latest patched version of TensorFlow-GPU or apply the provided workaround to mitigate this vulnerability. For further support on vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow-gpu →

FAQ