UTCS Colloquia - Isil Dillig, Faculty Candidate, College of William and Mary, "Introspective Static Analysis via Abductive Inference" ACE 2.302

Contact Name: 
Kate Callard
Location: 
ACES 2.302
Date: 
Mar 5, 2013 11:00am - 12:00pm

Signup Schedule: http://apps.cs.utexas.edu/talkschedules/cgi/list_events.cgi

Talk Audience: UTCS Faculty, Grads, Undergrads, Other Interested Parties

Host: Keshav K Pingali 

Talk Abstract: When software verification tools fail to prove the correctness a program, there are two possibilities: Either the program is indeed buggy  or the warning generated by the analysis is a false alarm. In this situation, the burden is on the programmer to manually inspect the warning and decide whether the program contains a real bug. However, this task is quite error-prone and time consuming and remains an impediment to the adoption of static analysis in the real world. In this talk, we present a new "introspective" approach to static analysis that helps programmers understand and diagnose error reports generated by the analysis. When our analysis fails to verify the program, it can meaningfully interact with users by generating small and relevant queries that capture exactly the information the analysis is missing to validate or refute the existence of an error in the program. The inference of such missing facts is an instance of logical abduction, and we present a new algorithm for performing abductive inference in logics that admit quantifier elimination. Since our abduction algorithm computes logically weakest solutions with as few variables as possible, the queries our technique presents to the user capture the minimum information needed to diagnose the error report. A user study we performed involving 56 programmers hired through ODesk indicates that our new technique dramatically improves the usefulness of static analysis to programmers.

Speaker Bio: Isil Dillig is currently an assistant professor of computer science at the College of William & Mary. She received her PhD in computer science from Stanford University in 2012. Isil also holds bachelors and masters degrees in computer science, also from Stanford University. Her research interests include program analysis, programming languages, formal methods, and automated logical reasoning.

Tags: