Good enough versus high assurance software testing and analysis methods

Abstract

Good enough evaluation methods are reviewed, including low volume statistical testing. High assurance software is viewed as requiring high volume statistical testing, or analysis methods that prove that a program has certain desired properties. The weaknesses of some previous high assurance strategies are discussed, and a new approach called pattern directed testing and analysis is described. In this approach, design patterns are used to give high assurance for critical generic properties, and enough assurance for problem dependent functionality.