Modeling the effects of combining diverse software fault detectiontechniques

摘要

Considers what happens when several different fault-findingntechniques are used together. The effectiveness of such multi-techniquenapproaches depends upon a quite subtle interplay between theirnindividual efficacies. The modeling tool we use to study this problem isnclosely related to earlier work on software design diversity whichnshowed that it would be unreasonable even to expect software versionsnthat were developed truly independently to fail independently of onenanother. The key idea was a “difficulty function” over theninput space. Later work extended these ideas to introduce a notion ofn“forced” diversity. In this paper, we show that many ofnthese results for design diversity have counterparts in diverse faultndetection in a single software version. We define measures ofnfault-finding effectiveness and diversity, and show how these might benused to give guidance for the optimal application of differentnfault-finding procedures to a particular program. The effects onnreliability of repeated applications of a particular fault-findingnprocedure are not statistically independent; such an incorrectnassumption of independence will always give results that are toonoptimistic. For diverse fault-finding procedures, it is possible forneffectiveness to be even greater than it would be under an assumption ofnstatistical independence. Diversity of fault-finding procedures is angood thing and should be applied as widely as possible. The model isnillustrated using some data from an experimental investigation intondiverse fault-finding on a railway signalling application