Skip to content
← All stories

Cloud Security · Identity · AI Security

Hackers abuse AntiFox AI tool to steal Amazon AWS credentials

Attackers have used the AntiFox AI pentesting platform to break into web servers and steal AWS credentials, letting them access cloud secrets and data. Amazon says its own systems were not breached.

Ross Grant3 min read

AI chip glowing inside a circuit board

The short version

Hackers have used a tool called AntiFox, which is meant for security testing, to break into web servers and steal passwords for Amazon cloud accounts. With these passwords, they could access and take sensitive information stored in the cloud. Amazon says its own systems were not hacked, but some customers' accounts were accessed using stolen credentials. Security experts recommend checking for unusual activity and changing passwords if you think your account was affected.

What happened?

Attackers used the AntiFox AI pentesting platform to break into Microsoft IIS web servers and steal Amazon AWS credentials, according to multiple security researchers. The attackers automated their intrusions using AntiFox, gaining access to sensitive cloud secrets and data. Amazon says its own systems were not breached, and that attackers used credentials stolen from customers' applications.

How the attack worked

Security researchers at Arctic Wolf have observed attackers using AntiFox, an open-source AI-driven penetration testing tool, to break into internet-facing Microsoft IIS web servers and Amazon Web Services (AWS) accounts. The attackers hosted their own AntiFox instances on Alibaba Cloud, then used the tool to automate steps like scanning for vulnerabilities, deploying web shells (malicious scripts that let them control servers), and searching for sensitive files. In some cases, they gained access to servers by exploiting unknown vulnerabilities in public IIS applications. In others, they started with valid AWS credentials, though it is not clear how these were obtained.

Once inside a compromised IIS server, the attackers used AntiFox to plant an ASPX web shell and run commands to gather information about the system. They looked for configuration files, database connection strings, and application settings. Using credentials found in these files, they tried to connect to internal production databases. The attackers also copied configuration and log files into folders accessible from the web, making it easier to steal them.

On the AWS side, the attackers used valid IAM (Identity and Access Management) users and EC2 roles to make hundreds of API requests in a short period. They cycled through multiple identities, probing for permissions and trying to access services like S3 (storage), Secrets Manager, Lambda, and more. Most requests were blocked, but some succeeded, letting the attackers access production secrets and, in some cases, read, copy, or delete S3 objects.

What was taken

The main target was AWS credentials and secrets stored in cloud environments. In at least one case, the attackers successfully retrieved a production secret from AWS Secrets Manager. They also accessed and manipulated objects in S3 storage. On compromised IIS servers, they stole configuration files, database connection details, and log files, which could be used for further attacks or to move deeper into the network.

The response

Amazon has confirmed that its own AWS services were not compromised. Instead, attackers used credentials stolen from customers' applications to access those customers' AWS accounts. Amazon recommends following best practices: use least-privilege permissions, require IMDSv2 on EC2 instances, avoid storing long-term access keys in application files, and enable Amazon GuardDuty for threat detection. If suspicious activity is detected, Amazon advises revoking affected sessions and rotating credentials immediately.

Researchers recommend monitoring for unusual web shell activity on servers, rapid or unusual AWS API calls from unfamiliar infrastructure, and reviewing logs for suspicious access. Organisations should patch public-facing applications, restrict application permissions, and enable detailed logging to detect unauthorised changes.

What you should do

If you run web servers or AWS environments, review your systems for signs of compromise. Look for new or suspicious .aspx files, unexpected command execution by IIS processes, and bursts of AWS API requests from unusual locations. Rotate any credentials that may have been exposed, and enforce least-privilege access. Enable logging and monitoring to catch similar attacks early. Prioritise patching and securing internet-facing applications, especially those that handle file uploads or process documents.

Claimed vs confirmed

  • Confirmed

    Attackers used AntiFox to steal AWS credentials and access cloud secrets.

    Multiple security researchers (Arctic Wolf, reported by BleepingComputer, The Hacker News, Security Affairs, TechRadar) observed and documented these attacks.

  • Denied

    Amazon AWS services themselves were breached.

    Amazon stated to BleepingComputer that AWS services were not compromised; attackers used credentials stolen from customer applications.

Why it matters

This attack shows how AI-powered security tools can be misused by criminals to automate and speed up attacks. It highlights the risk of storing sensitive credentials in application files and the importance of monitoring for unusual cloud activity. Organisations using cloud services need to review their security practices to prevent similar breaches.

Key points

  1. Attackers used AntiFox, an AI pentesting tool, to automate intrusions.
  2. Valid AWS credentials were used to access cloud secrets and data.
  3. Amazon says its own systems were not breached; customer credentials were misused.
  4. Most attack steps were automated, making detection harder.
  5. Security teams should monitor for web shells and unusual AWS activity.

Sources

  1. BleepingComputer

    Reference

  2. The Hacker News

    Reference

  3. Security Affairs

    Reference

  4. TechRadar

    Reference

  5. Dark Reading

    Reference

The day's cyber news in 5 minutes.

Cyberend Daily is one short email with the stories that matter, written so anyone can follow.

Includes our free beginner roadmap · No spam · Unsubscribe anytime