Cybersecurity researchers at Arctic Wolf detected attackers using infrastructure hosted on Alibaba Cloud to run self-hosted copies of AntiFox, an open-source penetration testing platform powered by AI clients. This infrastructure was used to launch multiple malicious activities, including vulnerability exploitation, deploying web shells, credential validation, system configuration discovery, and stealing secrets from Amazon Web Services (AWS) environments.
AntiFox is originally a fork derived from the testing tool (ARTEX 0.1.1), featuring a Chinese-language user interface that allows users to deploy AI agents to automate routine scanning tasks, such as mapping exposed services and scanning endpoints for security vulnerabilities. Security observations indicated that attackers used native IAM identity privileges and EC2 instance roles to execute authenticated requests against Amazon APIs.
Nature of the detected attacks
In one observed case, attackers deployed a malicious ASPX web shell to an internet-exposed IIS server, enabling them to execute operating system commands, access application settings and database connection strings, and copy configuration files and logs to web-accessible folders.
In another separate incident, the attackers used a valid AWS identity to read, copy, and delete S3 storage objects. In other attempts targeting endpoints, attackers navigated through multiple identities and sent hundreds of programmatic requests within thirty minutes. Although security systems successfully blocked most of these attempts, the attackers managed in some cases to extract sensitive production secrets.
What changes practically?
This development highlights the growing exploitation of AI-driven penetration testing tools and open-source frameworks in sophisticated automated attacks, similar to what was recently observed with the ARTEX AI tool that targeted banks in South Korea and cyber espionage campaigns linked to groups such as Vapor Panda.
Although AntiFox developers explicitly prohibit using the tool against live systems, in real-world penetration testing, or for illegal activities, the platform does not include strict technical controls to prevent abuse on cloud hosting platforms.
Protection and defense recommendations
Researchers provided technical guidelines for system administrators and defense teams, including:
- Correlating and tracking malicious web file activity on IIS servers with unusual access to AWS services.
- Reviewing logs and successful API requests that coincide with waves of security denials.
- Revoking affected sessions and keys in the AWS cloud immediately and rotating compromised credentials.