Automated Fitness Guided Fault Localization
Department
Computer Science
Major
Computer Science
Research Advisor
Tauritz, Daniel R.
Advisor's Department
Computer Science
Funding Source
Missouri S&T Opportunities for Undergraduate Research Experiences (OURE) Program
Abstract
For a given computer program, logical flaws produce invalid results which can be difficult, and thus costly, to correct. Finding software errors is a significant step towards fixing them. Given this, having a tool to automatically identify the most likely parts of the program containing these faults would be a huge advantage. Classical fault localization tools require an “oracle,” typically a human expert, to determine if the program is working as intended. The research presented here novelly employs a fitness function which automates the process by computationally enforcing the program specifications. The current Fitness Guided Fault Localization (FGFL) system combines two techniques: (1) trace comparison, and (2) trend-based line suspicion. Two additional techniques are under investigation, but require additional research to overcome their limitations, called: (1) run-time fitness monitor, and (2) critical slicing. Empirical comparisons are statistically analyzed for significance.
Biography
Alex is from Dorsey, Illinois. He is currently a sophomore in Computer Science, an Undergraduate Research Assistant in the Natural Computation Laboratory, a tutor for the Introduction to C++ labs (CS54), the Secretary for the Missouri S&T Association for Computing Machinery (ACM) Student Chapter, and will be interning this summer in the Center for Cyber Defenders at Sandia National Laboratories.
Research Category
Sciences
Presentation Type
Poster Presentation
Document Type
Poster
Location
Upper Atrium/Hallway
Presentation Date
10 Apr 2012, 9:00 am - 11:45 am
Automated Fitness Guided Fault Localization
Upper Atrium/Hallway
For a given computer program, logical flaws produce invalid results which can be difficult, and thus costly, to correct. Finding software errors is a significant step towards fixing them. Given this, having a tool to automatically identify the most likely parts of the program containing these faults would be a huge advantage. Classical fault localization tools require an “oracle,” typically a human expert, to determine if the program is working as intended. The research presented here novelly employs a fitness function which automates the process by computationally enforcing the program specifications. The current Fitness Guided Fault Localization (FGFL) system combines two techniques: (1) trace comparison, and (2) trend-based line suspicion. Two additional techniques are under investigation, but require additional research to overcome their limitations, called: (1) run-time fitness monitor, and (2) critical slicing. Empirical comparisons are statistically analyzed for significance.
