SHF: Small: Collaborative Research: Diversity and Feedback in Random Testing for Systems Software

University Of Utah, Salt Lake City UT

Investigators

Abstract

Testing is an extremely important part of any software development effort, especially for programs that have access to sensitive resources (personal information, sensor data, etc.) and that can be reached through the Internet. The PIs' work will improve the state of the art in software testing for mobile applications running on the open source Android platform, resulting in fewer bugs and reduced testing effort. The foundation of the PIs' work is random testing, where random numbers are used as inputs to an algorithm for constructing test cases. Although random testing has been shown to be highly effective for discovering serious bugs in complex software systems, it suffers from various problems including the fact that it is very difficult to engineer a random tester that doesn't spend a lot of time re-exploring the same application behaviors over and over again. The PIs will build upon their "swarm testing" work, which has been shown to be an inexpensive way to increase the diversity of random test cases, and also to increase their effectiveness in discovering bugs. Additionally, the PIs are investigating how to marry random testing with modern symbolic execution methods, and how to use feedback from executions of the software under test in order to improve the efficacy of random testing. The development of more efficient and effective testing techniques and tools will lower the cost and raise the quality of software. Test coverage is a challenging, open problem that is being addressed here in a novel way.