Research On The Method Of Synthesizing The Credibility Evidence Of Safety-critical Software

Safety-critical software plays a very important role in the fields of aerospace,rail transit,industrial control,etc.Once failure occurs,it will have unpredictable consequences.Therefore,it is a very important and urgent problem to know the trustworthy of software.At present,the methods for evaluating the trustworthy of software based on evidence synthesis include the weighted average method,function method,fuzzy inference,and evidence theory.The weighted average method is simple and intuitive,but it can't show the relationship between the trusted attributes of the software.The function method can reflect the relationship between software attributes to some extent,but it can't guarantee the objectivity of results because this method mainly depends on specialist marking methods when collecting evidences.The DS evidence theory can reduce the subjective factors in evaluation process to some extent,but there are some problems such as evidence synthesis paradox and attribute relevance in the fusion process.By analyzing the advantages and disadvantages of the existing algorithms which evaluate the trustworthy of software and according to the characteristics of metrics at different levels in the trustworthy evaluation model,this paper proposes a software trustworthy evaluation method based on layered evidence synthesis to evaluate the software trustworthy.This method combine the DS evidence theory and function method,collecting the trustworthy evidences in software life cycle,correcting the high conflict of evidences to reduce them,and improving the method of evidences corrected by judging the conflicting degrees of evidences firstly and correcting them according to the degree of conflict.On the basis of this,considering the independent of credible evidences,we use DS evidence theory to fuse at the credible evidence level.At the attribute layer,considering the correlation among the trustworthy attributes,it is impossible to use DS evidence theory.This paper analyzes the five construction criteria of the software trustworthy measurement model,and evaluates the attribute-level metrics through the trustworthy metric model to ensure the objective accuracy of the final evaluation result to some extent.Finally,through the detailed analysis of the actual case of the "railway Point Monitoring" system and comparing the experimental results with other trustworthy evaluate methods shows that the validity and rationality of our proposed method.