Research Of Neglected Conditions Defects Discovery Method Based On Graph Data Mining

With the rapid development of the software industry, the little defects in the softwarealmost used by malicious attackers, and cause huge loss to enterprise and ordinary users.Therefore, the problem of safety in the software is more and more important.In recent years, some methods combined Data Mining with Software Testing, and thisbecome a kind of new software test ideas. In this paper will finish a software testingsystem using Mining Maximal Frequent Subgraphs. The Program Dependence Graph setsas the inputs, and mining the correct rules in software, finally, with these rules to verifythe correctness of the program.In this paper, the defects discovery method can be divided into three process, datapreprocessing, rule mining and defects matching. This paper introduces the key problemsin the process of each step, divided data preprocessing into five steps and describedrespectively. Approach a rule mining algorithm based on the relationship between closedfrequent subgraphs and maximal frequent subgraphs called MCFSM, prove that MCFSMalgorithm can be better adapt to the mining of PDG in programs compare to other similaralgorithms. Matching the rules with the data after preprocess, for every rule looking forthe match program code from the input data, finally output the matching results.The last of paper proposed a system according to the theory, introduce the functionmodule and the design of the system. The experiments show that, the method proposed inthis paper is effective, feasible and acceptable.