An insight of double-faults interactions in program: An empirical study

Abstract

The study of multiple faults is becoming a hot spot. In the large software program, Multiple faults may interact with each other in some ways. In this paper, we investigated the property of multi-faults through the method of fault-injection in Siemens suits. Inspired by previous research, we focused our attention on three types of fault interaction: independent faults; faults masking; faults construction and explored what factors lead their interference in nature. The empirical study in Siemens suits showed that: 1). the probability of two faults interference is less than 1%, which means the independent assumption holds true in most cases. 2). faults masking is more frequent than the fault construction. 3). the occurrence of faults interference is not random. By focusing on and analyzing the double-faults versions that occur high-frequency interference, we find they always involve the same variable, which means the occurrence of fault interference have a certain condition.