基于函数调用关系分析软件网络特征

发布时间：2018-09-12 07:49

【摘要】：互联网的飞速发展在方便社会的同时,也带来了一系列的网络隐患。针对软件系统稳定性和安全性的问题,本文基于软件网络系统节点之间的调用关系、调用顺序以及内在的调用次数,引入复杂网络理论知识,从网络拓扑结构的角度,对软件系统的节点、路径、社团组织等结构特征进行了研究和分析。首先,为了更清晰地展示软件执行过程中的相关信息,根据函数间的调用关系,提出一种构建软件网络模型的新方法。基于在动态软件执行过程中设定不同测试用例,多次执行软件系统,追踪执行过程中函数调用的相关信息。通过对多次执行结果进行合并,最终实现软件网络模型的构建。其次,在软件系统动态执行过程中,基于函数之间的调用关系、调用顺序与调用次数,分析直接相邻的邻居节点和可达节点的影响力,在算法CIWN(calInitialWeightOfNode)和算法CFWN(calFinalWeightOfNode)分别计算节点的初始权值NIW(Node Initial Weight)和最终权值NFW(Node Final Weight),也就是直接邻居的影响力和在某一范围内所有可达节点的影响力。并通过算法MTKN(MinTopKNodes)对节点的权值进行排序,挖掘网络中的关键节点。展示了关键节点在软件各版本中的分布情况,从而了解软件演化过程中关键节点的特点。再次,针对软件各部件之间的调用关系,提出一种挖掘算法IIEP(idenImporExePath),挖掘网络中的关键路径。算法考虑了每条路径总节点中关键节点所占百分比,分析软件不同版本中的重要路径,用以指导将来的软件维护和预测。最后,引入社团理论,依据软件网络中的关键节点,提出算法COC(createOriginalCommunity)初始化软件网络中的社团结构。基于节点间不同的依赖程度,给出算法ETC(expandTheCommunity)扩展初始社团,将其他节点划分到不同的社团结构中,获取最优的划分。考虑软件“高内聚,低耦合”的特性,采用标准模块度Q来评估软件网络中社团结构。本文通过对三款软件进行实验,分析软件的网络结构,挖掘软件系统的潜在性质,验证所提方法的有效性和正确性。
[Abstract]:The rapid development of the Internet not only facilitates the society, but also brings a series of network hidden dangers. Aiming at the problem of the stability and security of software system, this paper introduces the complex network theory knowledge based on the call relation between nodes, the order of calling and the number of internal calls, from the point of view of network topology. The structural features of the software system, such as nodes, paths and community organizations, are studied and analyzed. Firstly, in order to show the relevant information in the process of software execution more clearly, a new method of constructing software network model is proposed according to the call relation between functions. Based on setting different test cases in the process of dynamic software execution, the software system is executed many times, and the relevant information of function call during execution is tracked. Finally, the software network model is constructed by merging the multiple execution results. Secondly, in the process of dynamic execution of software system, based on the call relation between functions, the order and times of call, the influence of neighbor node and reachable node directly adjacent to each other is analyzed. In the algorithm CIWN (calInitialWeightOfNode) and the algorithm CFWN (calFinalWeightOfNode), the initial weight NIW (Node Initial Weight) and the final weight NFW (Node Final Weight), are calculated respectively, that is, the influence of the direct neighbor and the influence of all reachable nodes in a certain range. The weights of nodes are sorted by MTKN (MinTopKNodes) algorithm, and the key nodes in the network are mined. The distribution of key nodes in each version of the software is shown, so as to understand the characteristics of the key nodes in the process of software evolution. Thirdly, a mining algorithm named IIEP (idenImporExePath), is proposed to mine the critical paths in the network. The algorithm takes into account the percentage of key nodes in the total node of each path and analyzes the important paths in different versions of the software to guide future software maintenance and prediction. Finally, based on the key nodes in the software network, an algorithm named COC (createOriginalCommunity) is proposed to initialize the community structure in the software network. Based on the different degree of dependency between nodes, an algorithm called ETC (expandTheCommunity) is proposed to extend the initial community and divide the other nodes into different community structures to obtain the optimal partition. Considering the characteristics of "high cohesion and low coupling", the standard modularity Q is used to evaluate the community structure in the software network. In this paper, through the experiment of three kinds of software, the network structure of the software is analyzed, the potential properties of the software system are excavated, and the validity and correctness of the proposed method are verified.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.5;O157.5

【相似文献】