Remote Code Execution in MLflow Due to SHAP Load Explainer Flaw
CVE-2026-118627
7.8HIGH
What is CVE-2026-118627?
A vulnerability in MLflow's handling of SHAP explainer models permits attackers to execute arbitrary code on installations by exploiting unvalidated user-supplied data. This issue arises from improper validation during the loading process of SHAP models. Attackers must lure users to visit a malicious page or open a compromised file to successfully initiate the exploit, which can lead to significant security risks if left unaddressed.
Affected Version(s)
MLflow 3.11.2.dev0
