AI-Agent E-Commerce Skimming Campaign

AI-Agent Campaign E-Commerce Skimming Magecart-style Injection Credential Theft 600K+ Cards

Overview

A financially motivated threat actor reportedly used multiple open-source AI-agent frameworks to automate reconnaissance, exploitation, credential theft, lateral movement, and payment-card skimmer deployment against online retailers.

Public reporting indicates more than 600,000 payment-card records were stolen, while skimming code was deployed across 100+ websites.

Technical Details

Three AI-agent tools were reportedly used:

• Strix for automated reconnaissance and vulnerability scanning.
• Cairn for exploitation and post-compromise activity.
• Hermes for orchestration and campaign management.

Reported techniques included:

• SQL injection
• MFA bypass
• Web-shell deployment
• Privilege escalation
• Credential theft
• Database extraction
• CDN/object-storage manipulation
• Kubernetes configuration changes
• Malicious JavaScript injection into checkout flows

Cron jobs were reportedly used to restore skimmers after removal, increasing persistence and recovery speed.

Technical Specifications

Attribute	Details
Threat Type	Financially motivated, automation-heavy intrusion and skimming campaign
Automation Stack	Strix (recon/scanning), Cairn (exploitation/post-compromise), Hermes (orchestration)
Initial Access Techniques	Web-application exploitation, including SQL injection
Post-Compromise Activity	Credential theft, privilege escalation, lateral movement, data extraction
Skimming Method	Magecart-style JavaScript injection in checkout infrastructure
Persistence Method	Cron-based restoration of removed skimming code

Affected Products

• Internet-facing e-commerce web applications and associated checkout systems.
• Retail cloud/application stacks with weak secret management or over-privileged service accounts.
• Environments where CDN/object storage and Kubernetes controls were insufficiently monitored or protected.

Attack Scenario

1. AI agents identify internet-facing targets.
2. Automated vulnerability scanning is conducted.
3. Web-application weaknesses are exploited.
4. Administrative access and/or web shells are established.
5. Credentials are stolen and databases/cloud secrets are accessed.
6. Payment-card information is extracted.
7. Magecart-style JavaScript skimmers are injected into checkout infrastructure.
8. Persistence is maintained and card-data collection continues.

Impact Assessment

Primary ImpactOperational RiskCriticality

• More than 600,000 payment-card records reportedly stolen
• 100+ websites infected with skimming code
• Sustained exposure to payment fraud and account abuse

• Automated offensive workflows increased campaign scale and speed
• In at least one reported case, autonomous cleanup activity deleted database tables
• Demonstrates potential for unintended destructive effects from autonomous agent operations

• High to Critical due to broad customer impact, payment-data exposure, and persistent skimmer reinfection behavior

Mitigation Strategies

1. Patch internet-facing web applications and eliminate SQL injection vulnerabilities.
2. Enforce strong MFA and monitor MFA-bypass attempts.
3. Apply least privilege to application, database, cloud, and CI/CD accounts.
4. Protect cloud secrets and rotate exposed credentials.
5. Monitor checkout JavaScript and third-party scripts for unauthorized changes.
6. Implement Content Security Policy (CSP) and Subresource Integrity (SRI) where appropriate.
7. Monitor CDN, object-storage, Kubernetes, and deployment-configuration changes.
8. Detect unexpected web shells, cron jobs, and unauthorized administrative accounts.
9. Use database activity monitoring and alert on large payment-data queries or exports.

Resources and References

Public Reporting

• Malicious AI agents steal 600K credit cards, infect 100+ sites with skimmers

---

Last Updated: September 29, 2026