CVE-2020-15199: TensorFlow - RaggedCountSparseOutput Validation Failure

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 data structures, including ragged tensors, which are tensors with non-uniform shapes. Ragged tensors are commonly used in natural language processing and other applications where data has variable lengths. However, in versions prior to 2.3.1, TensorFlow does not properly validate the input arguments for the RaggedCountSparseOutput function, leading to a vulnerability.

Vulnerability Detail

The vulnerability (CVE-2020-15199) arises from the lack of validation in the RaggedCountSparseOutput function of TensorFlow. Specifically, the code fails to check if the splits tensor, which is required for initializing a data structure called BatchedMap, has the minimum required number of elements. If the splits tensor is empty or has only one element, a SIGABRT signal is raised by the operating system, resulting in a crash. An attacker could potentially exploit this vulnerability to cause a denial-of-service (DoS) condition or disrupt the execution of TensorFlow-based applications.

pip install --upgrade "tensorflow==2.21.0"

pip install --upgrade "tensorflow==3cbb917b4714766030b28eba9fb41bb97ce9ee02"

Workaround

If upgrading to TensorFlow 2.3.1 is not immediately feasible, a possible workaround is to manually validate the splits tensor before passing it to the RaggedCountSparseOutput function. This can be done by checking the number of elements in the splits tensor and ensuring it meets the minimum requirement. Here's an example of how to perform this validation in Python:

import tensorflow as tf

splits = ...  # The splits tensor

if tf.size(splits) < 2:
    raise ValueError('The splits tensor must have at least two elements.')

# Proceed with using the `RaggedCountSparseOutput` function
...```

Conclusion

In conclusion, the vulnerability (CVE-2020-15199) in TensorFlow versions prior to 2.3.1 allows for a validation failure in the RaggedCountSparseOutput function. This can be exploited to cause a DoS condition or disrupt TensorFlow-based applications. Users should update to TensorFlow 2.3.1 or later to mitigate this vulnerability. If immediate upgrade is not feasible, manual validation of the splits tensor can serve as a temporary workaround. For further assistance or support in vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow →

FAQ