AntiFox AI Pentest Tool Pulled a Live AWS Production Secret

AntiFox, an open-source AI pentesting platform forked from ARTEX, has been linked to real intrusions that reached a production secret inside a victim's AWS account, The Hacker News reported based on Arctic Wolf Labs research. Investigators traced the activity to Alibaba Cloud systems that were also hosting AntiFox interfaces, chaining an ASPX web shell on an internet-facing IIS server to credential harvesting and then to cloud reconnaissance.

Inside AWS, the agentic framework enumerated IAM, EC2, S3, Lambda, Secrets Manager, KMS, RDS, ECS, EKS, Redshift, Athena, Glue, Systems Manager, API Gateway, CloudWatch, CloudFormation, DynamoDB, and Amazon Bedrock using List, Get, and Describe calls. Most requests returned AccessDenied, but the sequence included a "successful GetSecretValue request against a production secret." One intrusion lasted roughly half an hour and burned through hundreds of API calls while the agent cycled through multiple AWS identities.

The platform describes itself as built atop the "solid foundation" of ARTEX, with a "planner / worker / main-agent multi-agent system" running the full engagement over a dual-graph core, according to its README. On GitHub, the project has been forked 251 times and starred 474 times.

"Agentic frameworks such as these make penetration testing techniques more accessible to threat actors, while increasing the speed and consistency," Arctic Wolf wrote. The firm counted roughly 45 open-source agentic attack frameworks grouped into five categories, including network kill chain tools, web-app focused agents, and C2-integrated harnesses. It is the second AI-agent-plus-stolen-credentials story on our radar this week, after Railway's breach that racked up $740K in fraudulent AI charges on customer keys.

Arctic Wolf did not name the victim organization or disclose how the attackers first obtained the AWS credentials.