Research On Method For Mining Software Security Fault Feature Pattern

Different from the general hardware fault, the traditional fault analysis techniques can not play its effective role about the software fault. There exist many problems in the area of software security fault testing, such as the high false negative rate and false alarm rate,the poorness of automatically discovering knowledge and limitations of relatively large. Based on the above problem, the key point of this paper is the knowledge mining of the software security fault feature pattern, the improved software fault tree technique and sequential pattern mining technique are used to discuss how to mining the knowledge in the software security fault, and resolve the software security fault problems.Firstly, an automatic modeling algorithm of software fault tree is proposed for resolving the complexity of manual modeling algorithm. In the half part of the algorithm, it learns the compiler front-end technology, including mainly lexical analysis based on the state transition graph, recursively descent syntax analysis based on the grammar which is structured according to the top-down analysis strategy and syntax tree generation; in the other part there are the syntax tree extension based on the breadth traversal strategy and generation of software fault tree based on the depth traversal strategy.Secondly, for narrowing the scale of problem, a fast analysis algorithm is proposed, firstly, the logic simplification of the software fault tree based on the greedy strategy algorithm is implemented in the algorithm; then, the software fault tree is divided into modules based on the depth traversal; after that the cut-set matrix is used to expressed the modules and the process of solving is carried out on the cut-set matrix; the module results are synthesized in the last. In the end of the algorithm, the minimum cut-set about the whole software fault tree will be obtained, which can supply the data about software security fault for the mining of feature pattern knowledge of the software security fault in the next.Finally, a new closed sequential pattern mining algorithm based on positional data, FCSM-PD is proposed to obtain the software fault feature pattern, which can be used to construct the knowledge base of software fault, and guide the design process of software's source code. The algorithm can efficiently deal with the software fault data by the improved BI closure testing strategy, which greatly reduces the process time by narrowing the search space of sequence and decreasing the process of closure testing. The datasets with different characteristics are used to do the experiments for testing the algorithm FCSM-PD. By comparing and analysis the results of different treatments, we can see that the efficiency of the algorithm is better than the others.