CVE-2021-37641: Heap OOB in tensorflow

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. TensorFlow supports various operations and functions for manipulating tensors, including the RaggedGather function. This vulnerability affects versions of TensorFlow prior to 2.3.4, 2.4.3, and 2.5.1. The vulnerability allows an attacker to trigger a read from outside the bounds of heap allocated buffers, potentially leading to information disclosure or denial of service.

Vulnerability Detail

The vulnerability arises in the RaggedGather function of TensorFlow. If the arguments provided to this function do not determine a valid ragged tensor, an out-of-bounds read can occur. This can be exploited by an attacker to access sensitive information or crash the application. The issue is caused by a lack of proper input validation and boundary checks in the implementation of the function. The vulnerability has a CVSS score of 7.1, indicating a high severity.

pip install --upgrade "tensorflow==2.21.0"

Workaround

As a workaround, users can avoid using the RaggedGather function or ensure that the arguments provided always result in a valid ragged tensor. This can be achieved by validating the input data and ensuring that the indices used for gathering are within the bounds of the tensors. Here is an example of how to validate the input and avoid triggering the vulnerability:

import tensorflow as tf

params_nested_splits = [0, 0, 0]
params_dense_values = [1, 1]
indices = [0, 0, 9, 0, 0]

if all(split >= 0 for split in params_nested_splits) and all(index < len(params_dense_values) for index in indices):
    tf.raw_ops.RaggedGather(
        params_nested_splits=params_nested_splits,
        params_dense_values=params_dense_values,
        indices=indices,
        OUTPUT_RAGGED_RANK=0
    )
else:
    print('Invalid input for RaggedGather')

By performing these checks, the vulnerability can be mitigated.

Conclusion

The vulnerability in TensorFlow's RaggedGather function can be mitigated by updating to the patched versions of TensorFlow. Users should ensure that their TensorFlow installation is up to date to prevent potential information disclosure or denial of service. For further support and assistance with vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow →

FAQ