EXE - Automatically Generating Inputs of Death

Explore the groundbreaking EXE system for automated software testing through this conference talk that examines how concolic execution revolutionizes the discovery of critical software vulnerabilities. Learn about the innovative approach that combines concrete execution with symbolic analysis to automatically generate test inputs that can crash programs or expose security flaws. Discover how EXE balances the speed of random fuzzing with the precision of symbolic execution by running actual code on concrete inputs while simultaneously building logical constraints for unexplored program paths. Understand the technical implementation that instruments program execution to track symbolic constraints and selectively invokes SAT solvers only when needed to explore alternative execution branches. Examine how this hybrid methodology overcomes the limitations of pure symbolic execution, which can become computationally expensive on complex constraints, while maintaining the systematic path exploration capabilities that random testing lacks. Gain insights into the broader landscape of autonomous testing techniques and their role in complementing traditional human-written tests by leveraging computational power to uncover "unknown unknowns" in software behavior. The presentation covers the theoretical foundations, practical implementation challenges, and real-world applications of concolic execution in modern software security and reliability testing.