Software Network Models Based on Dynamic Execution for Fault Propagation Research

The growing complexity makes it difficult for researchers to quantify the behavior of software. By extracting the entities of software into nodes and the mutual effects between two entities into edges, a software can be transformed into a complex network which present both the topological structure and the entities´ properties. In this paper, software network models based on dynamic execution (SNMDE) are constructed for the purpose of studying fault propagation. First, the software are executed for hundreds of times, and each time of execution introduces an execution graph by recording the execution traces. Then the software network models are constructed by merging the execution graphs, the weights of the edges are calculated by combining the edges´ betweenness centralities and the execution conditions. Based on the proposed software network models, statistical analysis is conducted to show the complex network properties and to demonstrate the rationality of the network models.