Automation of Automation, or Fuzzing in the Age of GPT-4o & Gemini and Others
summary
The talk explains how to test large language models and machine-learning frameworks by using other AI models (e.g., GPT-4o and Gemini) to perform modern fuzzing. It demonstrates automating offensive and effective testing of learning systems, showing practical approaches to build AI-driven fuzzing pipelines for ML system validation.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:54Automation in vulnerability detection
- 8:18Developer and security collaboration
- 11:46Testing with LLMs
- 17:09Code vs test generation
- 22:54LLM testing methods
- 26:06SQLite and OpenSSL bugs
- 29:38LLM and software security
- 33:34TensorFlow and ML bugs
- 36:24Future and recommendations
- 40:17Summary
related talks
36mHow NOT to Train Your Hack Bot: Dos and Don'ts of Building Offensive GPTs
01h 24mOffensive Cybersecurity by AI Promises and Pitfalls
44mAI Foundations: Mitigate Risks and Boost SOC Efficiency
19mFuzzing in the GenAI Era
26mChallenging the Boundaries of Pentesting — The Age of AI
46mAutomated AI Security Testing Using Adversarially Fine-Tuned AI Agents
34mAI in Cybersecurity: Offense and Defense
49mChatGPT: A New Generation of Dynamic Machine Based Attacks?