A newly discovered Go-based botnet called NadMesh is actively scanning the internet for exposed AI services, harvesting cloud credentials and Kubernetes tokens from vulnerable deployments. The botnet, which surfaced in early July, has already collected thousands of AWS keys according to its operator’s own dashboard.
NadMesh uses Shodan integration to maintain a continuous queue of targets, specifically hunting for services running ComfyUI, Ollama, n8n, Open WebUI, Langflow, and Gradio. These tools are popular among AI developers and data scientists for tasks ranging from image generation to model hosting and workflow automation.
The targeting of AI infrastructure represents a significant trend in criminal computing. As organizations rapidly deploy AI services, many fail to properly secure them behind authentication and network controls. The exposed services become easy targets for botnets seeking compute resources, data, and network access.
The operator’s command dashboard reportedly displays 3,811 unique AWS keys collected through the scanning campaign. Each compromised credential can be used for data exfiltration, resource hijacking for cryptocurrency mining, or as a foothold for lateral movement within cloud environments.
Security researchers recommend that organizations running AI services ensure they are not exposed to the public internet without authentication, implement proper network segmentation, and monitor for unusual API usage patterns that could indicate compromise.
This article was adapted from The Hacker News. Read the original here.
