Title:   Mutation Analysis for Functional Defect Modeling
Authors: *Ian Harris and Qinshmang Zhang, UMass
Contact: harris@ecs.umass.edu

Abstract:
Design validation by simulation-based techniques is the most common
approach to verification due to the computational complexity of more
formal techniques. Several researchers have proposed different functional
fault models as tools in determining the degree of verification achieved
by a test sequence. As new fault models are proposed, it is essential
that the quality these fault models be measured before designers can
rely on the models for verification. The problem of measuring the
accuracy of a functional fault model for verification is analogous to the
problem of relating stuck-at fault coverage to defect coverage for 
manufacturing test. In both problems, an abstract high-level fault model
must be evaluated by comparison to a low-level defect model. The defect
model is tied directly to the source of the defects, while the fault
model is an abstraction of the behavior caused by defects. A functional
defect model is difficult to create because the source of design defects
is a human designer rather than a physical environment. We outline the
requirements of an accurate functional defect model, and we have developed
an initial defect model which is based on previous research in mutation
analysis for software testing. We demonstrate the utility of the
functional defect model in analyzing a simple functional fault model.