CVE-2020-15208: 
Python vulnerability analysis and mitigation

CVE-2020-15208 affects tensorflow-lite versions before 1.15.4, 2.0.3, 2.1.2, 2.2.1, and 2.3.1. The vulnerability was discovered by members of the Aivul Team from Qihoo 360 and was patched in TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, and 2.3.1. The issue involves a dimension mismatch vulnerability in TFLite's tensor handling functionality (GitHub Advisory).

The vulnerability occurs when determining the common dimension size of two tensors. TFLite uses a DCHECK statement which becomes a no-op outside of debug compilation modes. The function always returns the dimension of the first tensor, without properly validating against the second tensor's dimensions. The issue was fixed in commit 8ee24e7949a203d234489f9da2c5bf45a7d5157d by modifying the MatchingDim function to return the minimum of the two sizes (TF Commit). The vulnerability has received a CVSS v3.1 base score of 9.8 (CRITICAL) from NVD and 7.4 (HIGH) from GitHub (NVD).

When exploited, this vulnerability can result in reads and writes outside of bounds since the interpreter wrongly assumes there is enough data in both tensors. This can lead to data corruption and potential buffer overflow conditions (GitHub Advisory).

Malicious attackers can craft cases where the dimension of the first tensor is larger than that of the second tensor, leading to out-of-bounds operations. The vulnerability has been assigned CWE-125 (Out-of-bounds Read) and CWE-787 (Out-of-bounds Write) classifications (NVD).

Users are recommended to upgrade to TensorFlow versions 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1 which contain the fix. The patch modifies the MatchingDim function to return the minimum of the two tensor dimensions instead of always returning the first tensor's dimension (GitHub Advisory).