AI vs. AI: Microsoft Blocks Phishing Attack Using LLM-Generated Obfuscated Code

Microsoft Threat Intelligence reported a credential phishing campaign utilizing AI-generated code for obfuscation, which was ultimately blocked. The attackers used SVG files disguised as PDFs to hide malicious scripts, employing business terminology to mislead victims. The campaign, detected on August 18, involved a compromised email account and self-addressed messages to evade detection.

Microsoft identified several indicators of machine-generated obfuscation, including overly descriptive naming and complex code structure. Despite the sophistication of the attack, Microsoft Defender for Office 365 successfully detected and stopped it, highlighting that AI-enhanced threats can still be identified through specific artifacts.