Orion - Fuzzing Workflow Automation

Learn how to revolutionize software security testing through automated fuzzing workflows in this DEF CON 33 conference talk. Discover how Large Language Models (LLMs) can transform the traditionally manual and time-intensive fuzzing process by automatically identifying optimal fuzzing targets within codebases, generating precise test harnesses, and creating initial seed inputs. Explore the challenges software teams face when implementing fuzzing processes, including the significant time investment required for target identification and test harness creation, as well as the complex analysis and remediation of discovered vulnerabilities. Understand how this innovative automated solution addresses these pain points by leveraging LLMs for intelligent codebase analysis, streamlining the entire fuzzing workflow from target selection to bug reproduction and patch generation. Examine the demonstrated improvements achieved across all targeted areas of the fuzzing process, showcasing the practical effectiveness of integrating artificial intelligence and automatic code analysis into security testing methodologies. Gain insights into how this approach has successfully identified over 100,000 bugs across the industry while reducing the manual overhead traditionally associated with comprehensive fuzzing implementations.