AntiFox AI Pentesting Tool Linked to AWS Credential Theft

Researchers at Arctic Wolf Labs say attackers used Alibaba Cloud-hosted instances of AntiFox, an open source agentic penetration-testing framework, in intrusions that reached Microsoft IIS servers and Amazon Web Services environments.

The research report describes a chain that included web exploitation, web-shell deployment, configuration discovery, credential validation, cloud-service enumeration and secret access. In observed activity, valid AWS identities generated hundreds of cross-service API requests. Some discovery and secret-access requests reportedly succeeded, including access to a production secret.

The findings point to compromised customer credentials and applications, not a vulnerability in AWS itself. Arctic Wolf has not identified the initial IIS flaw in every case or attributed the activity to a named group, so the report establishes infrastructure and behavior rather than a definitive actor identity.

What defenders should watch

Security teams should correlate IIS worker processes spawning command shells or writing ASPX files with bursts of authenticated AWS API activity. A sequence containing many access denials followed by a successful secrets or storage request can be more revealing than either data source alone.

Organizations should rotate credentials exposed in application configuration, revoke affected sessions, require IMDSv2 for EC2 metadata access and apply least privilege to IAM identities. The larger concern is speed: agentic offensive tools can make a familiar sequence of post-exploitation steps easier to repeat and scale.