“The Magic of Statistics for Software Testing: How to Foresee the Unseen”

Ensuring software correctness is essential as software increasingly governs critical aspects of modern life, yet both formal verification and software testing face fundamental limitations. Testing, in particular, is inherently incomplete—it inevitably misses certain behaviors, leaving “unseen’’ program states and defects unaccounted for. This seminar introduces how statistical methods can help quantify these unseen, offering principled ways to assess what testing has not yet revealed. I will showcase recent advances that estimate residual risk, predict the reachability of critical behaviors, and extrapolate the future progress of fuzzing campaigns. By the end, participants will see how statistical thinking provides actionable insights that make testing more transparent, accountable, and empirically grounded.