CVE-2022-41896: TensorFlow GPU Crashes Due to Excessive Filterbank Channel Count in Mfcc Operation

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 open-source machine learning framework that provides a comprehensive ecosystem for building and deploying machine learning models. It supports various operations, including audio processing through the Mfcc (Mel-frequency cepstral coefficients) function. This function is widely used in speech and audio processing tasks. However, a vulnerability exists in the TensorFlow GPU package that can lead to crashes when the filterbank channel count exceeds the maximum allowed size.

Vulnerability Detail

The vulnerability arises from the ThreadUnsafeUnigramCandidateSampler in TensorFlow when the filterbank_channel_count parameter is set to a value greater than the allowed maximum size. This can lead to a crash of the TensorFlow application, disrupting any ongoing processes. The issue is triggered when the Mfcc operation is called with an excessively high filterbank channel count, which can be exploited by users who are unaware of the limitations.

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

Workaround

As a temporary workaround, avoid setting the filterbank_channel_count parameter to values greater than 2**31 - 1. Here is an example of how to safely call the Mfcc operation:

import tensorflow as tf

# Safe usage of Mfcc operation
try:
    tf.raw_ops.Mfcc(
        spectrogram=[[[1.38, 6.32, 5.75, 9.51]]],
        sample_rate=2,
        upper_frequency_limit=5.0,
        lower_frequency_limit=1.0,
        filterbank_channel_count=128,  # Use a safe value
        dct_coefficient_count=1
    )
except Exception as e:
    print(f'Error occurred: {e}')

Conclusion

The CVE-2022-41896 vulnerability in TensorFlow GPU can lead to application crashes when the filterbank channel count is set too high. Users are advised to update to the patched versions or implement workarounds to mitigate the risk. Regular monitoring for vulnerabilities using tools like Vulert can help maintain application security.

See every vulnerability recorded for tensorflow-gpu →

FAQ