CVE-2022-29200: Denial of Service Vulnerability in TensorFlow

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 developed by Google. It provides a comprehensive ecosystem of tools, libraries, and resources for building and deploying machine learning models. The vulnerability identified as CVE-2022-29200 affects TensorFlow versions 0 to 2.6.4, 2.7.0 to 2.7.2, and 2.8.0 to 2.8.1. This vulnerability can be exploited to cause a denial of service (DoS) attack by triggering a CHECK-failure in the tf.raw_ops.LSTMBlockCell function. The function does not properly validate the input arguments, leading to the failure and subsequent DoS.

Vulnerability Detail

The vulnerability arises from the lack of input argument validation in the implementation of the tf.raw_ops.LSTMBlockCell function in TensorFlow. Attackers can exploit this vulnerability by providing malicious input arguments that trigger a CHECK-failure. This can result in a DoS attack, rendering the affected TensorFlow application unresponsive. The vulnerability has a CVSS score of 5.5, indicating a moderate severity level. It was reported by Neophytos Christou from Secure Systems Lab at Brown University.

pip install --upgrade "tensorflow==2.21.0"

Workaround

As a workaround, users can implement their own input argument validation before calling the tf.raw_ops.LSTMBlockCell function. By ensuring that the input arguments have the correct ranks, users can mitigate the risk of triggering the CHECK-failure and potential DoS attack. Here is an example of how to validate the ranks of the input arguments in Python:

import tensorflow as tf

x = tf.constant(0.837607, shape=[28,29], dtype=tf.float32)
cs_prev = tf.constant(0, shape=[28,17], dtype=tf.float32)
h_prev = tf.constant(0.592631638, shape=[28,17], dtype=tf.float32)
w = tf.constant(0.887386262, shape=[46,68], dtype=tf.float32)
wci = tf.constant(0, shape=[], dtype=tf.float32)
wcf = tf.constant(0, shape=[17], dtype=tf.float32)
wco = tf.constant(0.592631638, shape=[28,17], dtype=tf.float32)
b = tf.constant(0.75259006, shape=[68], dtype=tf.float32)

if len(x.shape) != 2 or len(cs_prev.shape) != 2 or len(h_prev.shape) != 2 or len(w.shape) != 2 or len(wci.shape) != 0 or len(wcf.shape) != 1 or len(wco.shape) != 2 or len(b.shape) != 1:
    raise ValueError('Invalid ranks of input arguments')

# Call the LSTMBlockCell function
result = tf.raw_ops.LSTMBlockCell(x=x, cs_prev=cs_prev, h_prev=h_prev, w=w, wci=wci, wcf=wcf, wco=wco, b=b, forget_bias=1, cell_clip=0, use_peephole=False)

Conclusion

The vulnerability identified as CVE-2022-29200 in TensorFlow can lead to a denial of service (DoS) attack by exploiting a CHECK-failure in the tf.raw_ops.LSTMBlockCell function. It is recommended to update TensorFlow to versions 2.6.4, 2.7.2, or 2.8.1 to mitigate this vulnerability. Users can also implement input argument validation as a workaround. For further support on vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow →

FAQ