MLflow, FUXA Vulnerabilities Exploited for Cloud Credential Theft

Two critical vulnerabilities affecting the open-source artificial intelligence (AI) platform MLflow and the open-source SCADA/HMI software FUXA are currently being targeted by malicious actors. Independent security researchers from watchTowr and VulnCheck have reported observing active scanning and exploitation attempts against these flaws. The first vulnerability, identified as CVE-2024-29194, is a Server-Side Request Forgery (SSRF) flaw within MLflow. This SSRF vulnerability allows attackers to make unauthorized requests from the MLflow server to internal or external resources. Exploiting CVE-2024-29194 enables attackers to steal cloud credentials and sensitive secrets stored within the environment where MLflow is deployed. The researchers demonstrated that by chaining this SSRF vulnerability with other misconfigurations or vulnerabilities, attackers could potentially gain unauthorized access to cloud environments, including services like AWS, Azure, and Google Cloud. The implications of such an exploit are severe, as stolen cloud credentials can lead to data breaches, unauthorized resource usage, and significant financial losses. MLflow is widely used in the machine learning lifecycle for experiment tracking, model management, and deployment, making this vulnerability a significant concern for organizations leveraging AI technologies. The second vulnerability, CVE-2024-4058, affects FUXA, a web-based supervisory control and data acquisition (SCADA) and human-machine interface (HMI) software designed for operational technology (OT) and industrial automation. While specific details on the exploitation of CVE-2024-4058 are less elaborated in the initial reports, its presence in an OT system highlights the growing threat landscape for industrial control systems. Vulnerabilities in SCADA and HMI software can have devastating consequences, potentially leading to disruptions in critical infrastructure, manufacturing processes, and public utilities. The active exploitation of these vulnerabilities underscores the importance of timely patching and robust security practices for both AI development platforms and industrial control systems. Organizations using MLflow are strongly advised to update to the latest secure versions and review their security configurations to mitigate the risk of credential theft. Similarly, users of FUXA should apply any available security patches and assess their system's exposure. The watchTowr report specifically detailed how the MLflow SSRF vulnerability could be used to exfiltrate cloud provider metadata, which often contains sensitive access tokens and credentials. This highlights a sophisticated attack vector that leverages the interconnectedness of modern cloud-based AI development workflows. The exploitation of CVE-2024-29194 is particularly concerning as it bypasses traditional network perimeter defenses by originating requests from within the trusted MLflow environment. Security analysts emphasize that the widespread adoption of MLflow in research and production environments makes this a high-priority vulnerability for many organizations. The simultaneous reporting of exploitation for both MLflow and FUXA suggests a coordinated effort by threat actors to target diverse technology stacks, from AI development to critical industrial infrastructure.