CVE-2021-29608: TensorFlow CPU - Lack of Validation in tf.raw_ops.RaggedTensorToTensor

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 allows users to build and train neural networks, perform data preprocessing, and evaluate model performance. The package is widely used in various industries and research domains for a wide range of machine learning tasks.

Vulnerability Detail

The vulnerability (CVE-2021-29608) exists in the tf.raw_ops.RaggedTensorToTensor function of TensorFlow. Due to a lack of validation, an attacker can exploit an undefined behavior if input arguments are empty. The implementation of the function only checks that one of the tensors is not empty, but fails to validate the other ones. This oversight can lead to unexpected and potentially malicious behavior. Although there are some validations in place to prevent heap out-of-bounds access, these validations are disabled in release builds, rendering them ineffective. The vulnerability affects multiple versions of TensorFlow, including 2.1.4, 2.2.3, 2.3.3, 2.4.2, and prior versions. The TensorFlow team has addressed this issue and the fix will be included in TensorFlow 2.5.0.

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

Workaround

As a temporary workaround, users can manually validate the input arguments before calling the tf.raw_ops.RaggedTensorToTensor function. By ensuring that all input tensors are non-empty, users can mitigate the risk of triggering the undefined behavior. Here is an example of how to perform the validation in Python:

import tensorflow as tf

def safe_ragged_tensor_to_tensor(rt):
    if not rt.ragged_rank:
        raise ValueError('Input RaggedTensor must have ragged_rank > 0')
    if not rt.bounding_shape():
        raise ValueError('Input RaggedTensor must have non-empty bounding shape')
    return tf.raw_ops.RaggedTensorToTensor(rt)

Conclusion

The vulnerability (CVE-2021-29608) in TensorFlow's tf.raw_ops.RaggedTensorToTensor function can be exploited by an attacker if input arguments are empty. It is crucial for users to update their TensorFlow installations to version 2.5.0 or later once it is released to ensure the security and stability of their machine learning models and applications. In the meantime, users can apply a temporary workaround by manually validating the input arguments. For further support on vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow-cpu →

FAQ