CVE-2025-0649: 
TensorFlow vulnerability analysis and mitigation

Incorrect JSON input stringification in Google's Tensorflow serving versions up to 2.18.0 allows for potentially unbounded recursion leading to server crash. The vulnerability was discovered and disclosed in May 2025, affecting all versions of Tensorflow serving up to version 2.18.0 (NVD, Wiz).

The vulnerability stems from improper handling of JSON input stringification that can lead to unbounded recursion. The issue has been assigned a CVSS v4.0 base score of 8.9 HIGH with the vector string CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:H. The vulnerability has been classified under CWE-121 (Stack-based Buffer Overflow) (NVD).

When exploited, this vulnerability can cause a denial of service condition through server crashes. The high availability impact (VA:H) in the CVSS score indicates that the vulnerability can significantly affect system availability (Wiz).

The vulnerability has a network attack vector (AV:N) with low attack complexity (AC:L), requiring no privileges (PR:N) and no user interaction (UI:N). The EPSS (Exploitation Probability) score is reported at 5, indicating a relatively low likelihood of active exploitation (Wiz).

A fix has been implemented in the Tensorflow serving repository that improves the handling of large JSON requests by implementing iterative parsing and truncating large JSON strings to a fixed size when composing error messages (Github Commit).