CVE-2018-21233: 
Python vulnerability analysis and mitigation

TensorFlow versions before 1.7.0 contained a vulnerability in the BMP decoder component (CVE-2018-21233). The vulnerability was discovered by the Blade Team of Tencent and involves an integer overflow that causes an out-of-bounds read in the DecodeBmp feature of the BMP decoder located in core/kernels/decode_bmp_op.cc (TF Advisory).

The vulnerability stems from insufficient checking of header sizes and signed integer values in the BMP (bitmap image file graphics format) decoder. The issue received a CVSS v3.1 Base Score of 6.5 (MEDIUM) with a vector string of CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N, indicating network accessibility with user interaction required (NVD).

The vulnerability could lead to an unhandled process crash when processing invalid BMP files and potentially allow read access to unintended regions of the TensorFlow process memory, potentially exposing sensitive information (TF Advisory).

The vulnerability affects TensorFlow versions 1.3.0, 1.3.1, 1.4.0, 1.4.1, 1.5.0, 1.5.1, and 1.6.0. Exploitation requires the processing of a specially crafted BMP file by a vulnerable version of TensorFlow (TF Advisory).

The vulnerability was patched in TensorFlow version 1.7.0 and newer releases. A fix was also provided via GitHub commit 49f73c55, which implemented more stringent checks in the DecodeBmp function. Users running TensorFlow in production or processing untrusted data are encouraged to upgrade to version 1.7.0 or apply the security patch (TF Advisory).