Research On The Robustness Of Software Networks Based On Function Call

In recent years,the increasing complexity of software structure directly affects the stability of software system.Therefore,how to take effective measures to improve the stability of software network and make software quality more safe and reliable is of great significance.This paper takes real open source software as the research object.The main research contents are as follows:(1)Based on open source software Tar,My SQL and Linux source code analysis,builds the directed function call network model,the weak coupling network model of software system,the strong coupling network model and the strong connected subgraph model.The network structural properties such as degree distribution,clustering coefficient,average path length,network diameter,node dependence and influence are investigated.The experimental results show that all software networks exhibit the characteristics of "small-world" and "scale-free".The efficiency of function call in coupling network is higher than that of function call software network.The node dependence is of a positive correlation with the node's out-degree;the node influence is of a positive correlation with the node's in-degree;the node influence has a negative correlation with its dependence.(2)Based on the weak connectivity and strong connectivity robustness measure of directed networks,we use eight attack strategies to investigate the static robustness of software network.The experimental results show that,for each software network,high in-degree and high out-degree is the best attack strategy;under deliberate attack,the weakly coupled software network is more vulnerable.The network evaluated by strongly connected indicator is more fragile.The best attacking strategy is distinct when different robustness evaluation indicator is adopted.(3)Based on the Coupled map lattices(CML)fault model,this thesis introduces five node attack strategies to study the dynamic robustness of the software network,and adopts the relative damage scale indicators to measure.The experimental results show that,for each software network,the high betweeness is the best attack strategy.The vulnerability of directed function call software network,weakly coupled software network,strongly coupled network and strongly connected subgraph network increases successively.The research results can provide guidance for understanding the structure of software networks and improving the robustness of software systems.