Overview
Threat intelligence reports indicate active exploitation and scanning efforts targeting two significant open-source platforms: MLflow and FUXA. Threat actors are leveraging a critical Server-Side Request Forgery (SSRF) flaw in MLflow to compromise cloud credentials and secrets. Concurrently, malicious scanning is underway for a path traversal vulnerability in FUXA, an open-source SCADA/HMI software. Security teams should prioritize patching and monitoring to defend against these ongoing threats, according to The Hacker News.
Technical Analysis of Active Exploitation
The current threat landscape highlights two distinct but equally concerning vulnerability exploitation trends impacting critical operational and development environments.
MLflow SSRF Vulnerability (CVE-2026-64849)
A critical Server-Side Request Forgery (SSRF) vulnerability, identified as CVE-2026-64849, is being actively exploited in MLflow. MLflow is a widely used open-source platform for managing the machine learning lifecycle. According to watchTowr’s analysis, attackers are exploiting this flaw in MLflow’s model-registry webhooks to proxy requests through affected systems. This allows them to directly interact with internal services, specifically cloud metadata services, to exfiltrate sensitive cloud credentials and secrets.
WatchTowr detected indiscriminate scanning for exposed MLflow instances online almost immediately after the CVE was assigned on August 17, 2026. Evidence from their global honeypot telemetry confirms attackers are specifically targeting cloud-hosted MLflow systems. This exploitation aims to extract credentials and secrets from well-known internal IP addresses and services, highlighting a significant risk to cloud environments leveraging MLflow. The vulnerability is particularly dangerous because it bypasses prior security fixes due to its specific handling of web redirects. Security professionals researching how to detect MLflow cloud credential theft will find watchTowr’s observations crucial for identifying indicators of compromise related to this sophisticated SSRF exploitation.
FUXA Path Traversal Vulnerability (CVE-2026-25895)
Separately, malicious scanning efforts have been detected targeting a path traversal vulnerability in FUXA, an open-source, web-based SCADA/HMI software. VulnCheck observed scanning for CVE-2026-25895 starting August 18, 2026. A single IP address has been broadly scanning the internet for vulnerable FUXA instances, with approximately 60 FUXA installations currently exposed to the public internet.
The attacker’s request attempts to overwrite the main.js file with arbitrary data via the CVE-2026-25895 path traversal. While no Remote Code Execution (RCE) payloads have been observed dropped yet for this specific vulnerability, the intent to manipulate system files is clear. This activity follows a pattern, as other FUXA vulnerabilities, including CVE-2026-25939 and CVE-2023-33831, have seen active exploitation dating back to November 2025. Organizations should urgently consider FUXA CVE-2026-25895 patch guidance.
Mitigation and Recommendations
Organizations running MLflow or FUXA instances must take immediate action to mitigate these threats. Proactive measures are essential to prevent credential theft and system compromise.
- Prioritize Patching: For MLflow, apply patches for CVE-2026-64849 immediately. For FUXA, apply patches for CVE-2026-25895, as well as any outstanding patches for CVE-2026-25939 and CVE-2023-33831. Updating to the latest secure versions is the most effective defense against known vulnerabilities.
- Audit Logs: Thoroughly review audit logs for MLflow instances, especially those hosted in cloud environments. Look for signs of unusual outbound connections, attempts to access metadata services, or unexpected API calls that could indicate credential exfiltration.
- Credential Review: Conduct an immediate review to determine whether sensitive cloud credentials or secrets may have been exposed through MLflow instances. Rotate any compromised credentials without delay.
- Network Segmentation: Implement strong network segmentation for MLflow and FUXA deployments, especially for instances exposed to the internet. Restrict outbound connectivity from MLflow systems to only necessary services to limit the impact of SSRF exploits.
- Input Validation: Ensure all input to MLflow webhooks and FUXA interfaces is rigorously validated to prevent injection attacks and path traversals.
- Monitoring: Deploy continuous monitoring solutions to detect anomalous activity on MLflow and FUXA servers. This includes monitoring for unexpected file modifications, process execution, and network connections.
- Exposure Assessment: Identify and limit public exposure of MLflow and FUXA installations. Many instances of FUXA are unnecessarily internet-facing.
Adhering to these recommendations is crucial for mitigating MLflow CVE-2026-64849 SSRF risks and securing critical AI and OT infrastructure against ongoing exploitation efforts.
Related: Cloudflare Achieves FedRAMP High Status for Government, Dify AI Platform Vulnerabilities: How to Mitigate DifyTap Exploit