Misconfigured Open WebUI exposures exploited for AI-generated malware deployment
Attackers exploited a misconfigured and internet-exposed Open WebUI instance to upload and execute AI-generated malware that deployed cryptocurrency miners and credential-theft tools. Sysdig's security detection systems identified the attack in real time, and the incident highlights risks from widely-deployed AI tools lacking proper access controls.
What happened
Sysdig's Threat Research Team discovered attackers exploiting a misconfigured Open WebUI instance deployed by one of Sysdig's customers that was exposed to the internet with administrative privileges and no authentication. The attackers uploaded a malicious, obfuscated Python script through Open WebUI's plugin system, which allowed execution of commands on the exposed system.
The malicious script was AI-generated or heavily AI-assisted, containing inline format string variables and obfuscation via PyObfuscator. It downloaded and executed cryptocurrency miners targeting Monero and Ravencoin networks on Linux systems, and deployed components for credential theft, sandbox evasion and hardware discovery on Windows systems.
Reported impact
The attack deployed cryptominers on affected systems and established persistence through a systemd service masquerading as 'ptorch_updater.' On Windows, the payload included credential theft and hardware discovery capabilities alongside crypto mining functionality.
Index analysis
Sysdig reported that over 17,000 Open WebUI instances are currently exposed on the internet, though the number of misconfigured or vulnerable instances is unclear. The attack demonstrates a shift towards rapid malware development using generative AI tools, with attackers leveraging AI to generate obfuscated payloads.
Governance controls
Sysdig's runtime threat detection identified the threat in real time using YARA rules, behavioral detections and threat intelligence, detecting suspicious activity including unauthorized code compilation, domain lookups, and known miner communication protocols.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.