AntiFox AI agents used in AWS intrusions exposing production secrets
Original title: AntiFox AI Agents Linked to AWS Attacks That Exposed Production Secrets and Compromised Web Servers
Arctic Wolf saw an IIS ASPX webshell, hundreds of AWS API calls in about 30 minutes, and S3 read, copy, and delete.
Attackers are using AntiFox, a Chinese-language fork of ARTEX 0.1.1, against internet-facing web applications and AWS. Arctic Wolf linked self-managed instances on Alibaba Cloud to an ASPX webshell on Microsoft IIS that ran commands and copied configuration, logs, and database strings into a web-accessible directory. Valid IAM users and assumed EC2 roles were then used; most calls were denied, but some exposed production secrets. GitGuardian says few agent frameworks include secret scanners, and a lab ARTEX run found eight of nine planted credentials without using them.
Unconfirmed: The GitGuardian test measured discovery only and did not show the model using the credentials.
- The report followed Microsoft’s November 5 disclosure of ARTEX-assisted GitHub credential attacks.
- One case used multiple identities and generated hundreds of API requests in about 30 minutes.
- A separate incident read, copied, and deleted S3 objects.
- GitGuardian reviewed about 45 frameworks; only four had working secret scanners.
- The lab search began within six to 13 seconds and refused prompts asking for abuse.
Impact: Production secrets exposed; S3 objects read, copied, and deleted
- 1.Revoke affected AWS sessions and access keys and rotate exposed credentials.
- 2.Correlate IIS webshells with bursts of denied and successful AWS API calls.
- 3.Treat a deleted leaked file or commit as still valid until credentials are rotated.
- ATT&CK
- T1505.003 Web ShellT1059 Command and Scripting InterpreterT1552.001 Credentials In FilesT1078 Valid AccountsT1530 Data from Cloud Storage
- Organizations
- Arctic WolfGitGuardian
From the original article
2 min read at gbhackers.com
Attackers are using AntiFox, an agentic AI penetration-testing framework derived from ARTEX, to target internet-facing web applications and AWS environments. Reported on November 11, the findings follow Microsoft’s November 5 disclosure of ARTEX-assisted attacks against exposed GitHub credentials.
By Mayura Kathir · gbhackers.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.