AntiFox AI agents used in AWS credential theft and IIS webshell attacks
Attackers host AntiFox AI pentest agents on Alibaba Cloud to steal AWS credentials; rotate exposed keys and review AWS API activity.
- Published
- Updated
- Exploitation
- Yes, Arctic Wolf observed intrusions using AntiFox-linked infrastructure, as reported by BleepingComputer and GBHackers
What happened
Arctic Wolf reported that attackers are hosting self-managed instances of AntiFox, an open-source agentic penetration-testing platform, on Alibaba Cloud. They use it against web services and AWS environments, according to BleepingComputer. The hosting infrastructure was tied to vulnerability exploitation, webshell deployment, credential validation, configuration discovery and AWS secret harvesting (same source).
AntiFox began as a fork of ARTEX 0.1.1 and has a Chinese-language interface. It lets users deploy AI agents that automate tasks such as mapping exposed services and probing endpoints (BleepingComputer).
Reported details of the intrusions:
- In one case, attackers placed an ASPX webshell on an internet-exposed IIS server. They ran OS commands, read application settings and database connection strings, and copied configuration and log files into a web-accessible directory (BleepingComputer, GBHackers).
- The activity extended into AWS using valid identities, including IAM users and assumed EC2 roles. In one intrusion the attackers cycled through several identities and made hundreds of API requests in about 30 minutes. Most were denied, but some succeeded and yielded production secrets (BleepingComputer).
- In a separate incident, a valid AWS identity was used to read, copy and delete S3 objects (BleepingComputer).
- TechRadar adds that Arctic Wolf could not determine how the IIS servers were first compromised. It also reports attempts to use the harvested credentials to reach internal production databases (TechRadar). How the attackers obtained the AWS credentials in every case is not confirmed in the sources.
Amazon told BleepingComputer: “AWS services were not compromised and operated as designed.” It said the research describes actors using credentials taken from customers’ compromised applications (BleepingComputer).
BleepingComputer notes AntiFox remains on GitHub and has been updated to version 2.0. Its publishers prohibit use against live systems but describe no technical controls to enforce that.
Separate but related report. Elastic Security Labs described a different tool, CyberStrikeAI, not AntiFox. Between September 30 and October 16, 2026, one CyberStrikeAI server used credentials from eight AWS identities (five IAM users and three EC2 instance roles) across four accounts. One role read 56 EC2 user data scripts in 141 milliseconds. Elastic says it cannot prove an LLM agent made the decisions and does not know how the credentials were obtained (Elastic Security Labs). It notes that the instance-role sessions were used from an outside IP, which means the credentials were taken from the instances, for example through SSRF or host compromise. Do not read this as the same campaign as the AntiFox activity.
Am I affected?
The sources describe targeting of internet-facing web services, with IIS servers in the cases described, and of the AWS accounts linked to them. You may be exposed if:
- You run internet-facing IIS servers whose application files hold database connection strings or AWS keys.
- You keep long-term AWS access keys in application configuration, or EC2 roles are usable from instances that a webshell could reach.
- Your AWS logs show rapid requests across several services or identities, or access attempts from unfamiliar infrastructure (TechRadar).
Indicators such as specific IPs or domains for AntiFox are not confirmed in the sources provided. Elastic lists indicators for the CyberStrikeAI server it observed, and says none of its three servers was serving that interface as of November, so it recommends using them only for retrospective hunting.
What to do now
- Look for unfamiliar webshells and unexpected files on web servers, and review OS commands run without approval (TechRadar).
- Correlate IIS webshell activity with unusual AWS access, and review successful API calls inside bursts of denied requests (BleepingComputer, GBHackers).
- Revoke affected AWS sessions and access keys, and rotate credentials exposed through application files. (BleepingComputer, GBHackers). GBHackers, referring to Microsoft’s separate ARTEX investigation, adds that deleting a leaked file does not invalidate the credentials.
- Apply the AWS recommendations Amazon gave BleepingComputer:
- grant least-privilege permissions;
- require IMDSv2 on EC2 instances;
- keep long-term access keys out of application configurations;
- turn on Amazon GuardDuty.
- Enable detailed logging so suspicious access and unauthorized changes to stored data can be detected (TechRadar).
- If you suspect unauthorized activity, follow AWS guidance for remediating potentially compromised credentials or contact AWS Support (Amazon statement, via BleepingComputer).
- Consider that activity may be spread over time. Elastic’s CyberStrikeAI case included a 14-day gap, and its alerts were only meaningful once correlated by source IP (Elastic Security Labs).
Update, 14 November 2026
- BleepingComputer reports that ARTEX AI, the framework AntiFox was forked from, went closed-source after it was used in attacks on South Korean banks. (bleepingcomputer.com)
- BleepingComputer says CrowdStrike linked ARTEX to a cyberespionage campaign in which the Chinese hacking group Vapor Panda targeted more than 200 organizations across 14 countries. (bleepingcomputer.com)
Update, 17 November 2026
- TechRadar reports that Censys identified 1,465 hosts running exposed ARTEX consoles, the framework AntiFox was forked from. Roughly 10% of those hosts also ran at least one other tool commonly used in cyberattacks, and Censys found 32 publicly accessible directories on ARTEX hosts. (techradar.com)
Sources
- BleepingComputer: Hackers use AntiFox AI agents in AWS credential-theft attacks
- GBHackers: AntiFox AI Agents Linked to AWS Attacks That Exposed Production Secrets and Compromised Web Servers
- TechRadar: AntiFox AI agents hijacked by hackers to carry out theft of AWS credentials
- Elastic Security Labs: 56 AWS API calls in 141 milliseconds: Inside an AI-run intrusion