CVE-2021-29549: Denial of Service in TensorFlow

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 is widely used in various domains, including research, industry, and academia. It supports both CPU and GPU computations, allowing for efficient training and inference on different hardware architectures.

Vulnerability Detail

The vulnerability identified as CVE-2021-29549 affects the tf.raw_ops.QuantizedBatchNormWithGlobalNormalization function in TensorFlow. An attacker can exploit this vulnerability to cause a runtime division by zero error, leading to a denial of service. The vulnerability arises from a lack of validation in the implementation of the function, specifically in the computation of a modulo operation without checking if the divisor is zero. By manipulating the input shapes, an attacker can trigger scenarios where the divisor becomes zero, resulting in a division by zero error and subsequent denial of service. The impact of this vulnerability can disrupt the availability of TensorFlow-based applications and systems.

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

Workaround

As a temporary workaround, users can implement input validation in their own code to ensure that the input shapes passed to the tf.raw_ops.QuantizedBatchNormWithGlobalNormalization function do not trigger a division by zero scenario. This can be achieved by checking the input shapes and avoiding cases where the divisor becomes zero. Here is an example of how the input validation can be performed in Python:

import tensorflow as tf

def safe_quantized_batch_norm(input_tensor, scale, offset):
    if input_tensor.shape[0] == 0 or scale.shape[0] == 0 or offset.shape[0] == 0:
        raise ValueError('Invalid input shapes')
    else:
        return tf.raw_ops.QuantizedBatchNormWithGlobalNormalization(input_tensor, scale, offset)

Conclusion

In conclusion, the vulnerability identified as CVE-2021-29549 in TensorFlow can lead to a denial of service by causing a runtime division by zero error. It is crucial for users to update their TensorFlow installation to one of the fixed versions mentioned above to mitigate this vulnerability. Additionally, implementing input validation in user code can serve as a temporary workaround until the upgrade is performed. For further support and guidance on vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow-gpu →

FAQ