Date of Award

Spring 2016

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computing Sciences - (Ph.D.)

Department

Computer Science

First Advisor

Ali Mili

Second Advisor

Narain Gehani

Third Advisor

M. Rusinkiewicz

Fourth Advisor

Kurt Rohloff

Fifth Advisor

Thomas Wies

Sixth Advisor

Kazunori Sakamoto

Abstract

Over the last decade, research in verification and formal methods has been the subject of increased interest with the need of more secure and dependable software. At the heart of software dependability is the concept of software fault, defined in the literature as the adjudged or hypothesized cause of an error. This definition, which lacks precision, presents at least two challenges with regard to using formal methods: (1) Adjudging and hypothesizing are highly subjective human endeavors; (2) The concept of error is itself insufficiently defined, since it depends on a detailed characterization of correct system states at each stage of a computation (which is usually unavailable). In the process of defining what a software fault is, the concept of relative correctness, the property of a program to be more-correct than another with respect to a given specification, is discussed. Subsequently, a feature of a program is a fault (for a given specification) only because there exists an alternative to it that would make the program more-correct with respect to the specification. Furthermore, the implications and applications of relative correctness in various software engineering activities are explored. It is then illustrated that in many situations of software testing, fault removal and program repair, testing for relative correctness rather than absolute correctness leads to clearer conclusions and better outcomes. In particular, debugging without testing, a technique whereby, a fault can be removed from a program and the new program proven to be more-correct than the original, all without any testing (and its associated uncertainties/imperfections) is introduced. Given that there are orders of magnitude more incorrect programs than correct programs in use nowadays, this has the potential to expand the scope of proving methods significantly. Another technique, programming without refining, is also introduced. The most important advantage of program derivation by correctness enhancement is that it captures not only program construction from scratch, but also virtually all activities of software evolution. Given that nowadays most software is developed by evolving existing assets rather than producing new assets from scratch, the paradigm of software evolution by correctness enhancements stands to yield significant gains, if we can make it practical.

Share

COinS