Homoglyph-Based Attacks - Circumventing LLM Detectors

Learn about homoglyph-based attacks and techniques for bypassing LLM detection systems in this 31-minute conference presentation from DefCamp 2024. Delivered by Aldan Creo, Technology Research Specialist at Accenture Labs, at Eastern Europe's premier cybersecurity conference in Bucharest. Explore cutting-edge research on exploiting visual similarities between characters to circumvent AI detection mechanisms. Gain insights from this talk presented at an event that attracted over 2,000 cybersecurity professionals, researchers, and industry experts from 50 countries, continuing DefCamp's legacy as the region's most significant annual gathering for information security specialists since 2011.