CVE-2021-29585: Division by zero in padding computation in TFLite

Background

The TensorFlow Lite (TFLite) package is an open-source machine learning framework developed by TensorFlow. It is designed to run on resource-constrained devices, such as mobile and embedded devices, and provides efficient execution of machine learning models. TFLite includes various operations and functions for model inference, including padding computation for image processing. However, a vulnerability has been identified in the padding computation function of TFLite, which can lead to a division by zero error.

Vulnerability Detail

The vulnerability in TFLite arises from the ComputeOutSize function, which calculates the size of the output after padding. This function does not check whether the stride argument is zero before performing the division. As a result, if a user crafts a model that calls ComputeOutSize with a stride value of zero, a division by zero error will occur. This vulnerability can be exploited by an attacker to cause a denial-of-service condition or potentially execute arbitrary code. The impact of the vulnerability is considered low, with a CVSS score of 2.5.

pip install --upgrade "tensorflow==2.21.0"

Workaround

As a workaround, users can manually check the stride argument before calling the ComputeOutSize function to ensure it is not zero. If the stride value is zero, an alternative calculation or padding method should be used to avoid the division by zero error. Here is an example of how to perform the check in Python:

def compute_out_size(padding, image_size, filter_size, stride, dilation_rate=1):
    effective_filter_size = (filter_size - 1) * dilation_rate + 1
    if stride == 0:
        # Handle the case where stride is zero
        return 0
    elif padding == 'SAME':
        return (image_size + stride - 1) // stride
    elif padding == 'VALID':
        return (image_size + stride - effective_filter_size) // stride
    else:
        return 0

By explicitly checking the stride value, the division by zero error can be avoided.

Conclusion

In conclusion, the vulnerability in TFLite's padding computation function can result in a division by zero error. While the impact is considered low, it is important to update to the patched versions of TensorFlow once they are released. Users can also manually check the stride argument or use the provided workaround to mitigate the vulnerability. For further support on vulnerability remediation, please contact DevNack.com.

See every vulnerability recorded for tensorflow →

FAQ