Computer Science Distinguished Lecture
Date/Time: | Thursday, 07 Mar 2013 at 3:40 pm |
---|---|
Location: | 207 Marston Hall |
Cost: | Free |
Phone: | 515-294-6516 |
Channel: | College of Liberal Arts and Sciences |
Categories: | Lectures |
Actions: | Download iCal/vCal | Email Reminder |
Abstract
It is difficult to build complex systems that (almost) never go (badly) wrong, yet this is what we expect of airplanes and pacemakers and the phone system. In essence, we have to anticipate everything that could fail or go wrong, develop countermeasures, and then provide compelling evidence that we have done all this correctly. I will outline some of the intellectual challenges in construction of suitable evidence, particularly as applied to software. I will introduce the idea of "possibly perfect" software and its associated "probability of perfection" and describe how this relates to correctness and reliability. I will sketch some approaches to estimating a probability of perfection and touch on alternative proposals such as those based on "eliminative induction." I will then describe epistemic and logic uncertainties in high-assurance software and speculate on the relation between these and the notion of resilience.
Much of this material is based on Joint work with Bev Littlewood at City University UK. Some of it is described in a paper available at http://www.csl.sri.com/users/rushby/ abstracts/1oo2-tse