CVE-2021-29532: TensorFlow Heap Out-of-Bounds Read

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. It is widely used in various industries for tasks such as image recognition, natural language processing, and recommendation systems. However, a vulnerability has been discovered in TensorFlow that can be exploited by an attacker to force accesses outside the bounds of heap allocated arrays.

Vulnerability Detail

The vulnerability exists in the tf.raw_ops.RaggedCross function of TensorFlow. By passing in invalid tensor values, an attacker can trigger accesses outside the bounds of heap allocated arrays. This is due to a lack of validation in the implementation of the function. The code does not check if the index values used to access array elements are within the valid range, leading to out-of-bounds reads in the heap. An attacker can potentially leverage this vulnerability to read sensitive data or cause a denial-of-service condition. The issue has been identified and a fix will be included in TensorFlow 2.5.0. Additionally, the fix will be backported to TensorFlow 2.4.2, TensorFlow 2.3.3, TensorFlow 2.2.3, and TensorFlow 2.1.4, as these versions are also affected and still within the supported range.

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

Workaround

As a temporary workaround, users can implement input validation in their own code before passing tensors to the tf.raw_ops.RaggedCross function. By validating the tensor values and ensuring they are within the expected range, the risk of triggering the vulnerability can be mitigated. Here is an example of how input validation can be performed in Python:

import tensorflow as tf

def safe_ragged_cross(tensor):
    if not isinstance(tensor, tf.Tensor):
        raise ValueError('Input tensor must be a TensorFlow Tensor.')
    if tensor.dtype != tf.float32:
        raise ValueError('Input tensor must have dtype float32.')
    if tensor.shape[0] > 1000:
        raise ValueError('Input tensor size exceeds maximum limit.')
    # Perform safe RaggedCross operation
    return tf.raw_ops.RaggedCross(tensor)

Conclusion

In conclusion, a vulnerability has been discovered in TensorFlow that allows an attacker to perform out-of-bounds reads in heap allocated arrays. This can lead to unauthorized access to sensitive data or a denial-of-service condition. Users are advised to update to the latest patched version of TensorFlow or apply the respective patches for their affected versions. Implementing input validation in code can also serve as a temporary workaround. For further support on vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow-cpu →

FAQ