一种基于缺陷传播的复杂软件系统质量评估方法

With the background of large scale and high complexity for software systems, the identification of the key in software systems and the evaluation of their influence, and the quality assessment for software systems, can be benefit for selective monitoring, reducing the underlying risk of software maintains and improving software quality. Therefore, this proposal plans to identify and analyze the key vertices in software networks which can be extracted from source codes, based on the empirical studies on open-source software systems. Particularly, we will present the method of producing the propagating and propagated scopes on the basis of method-calling relations between the classes, then evaluate the propagating and propagated effects, and present the defect propagation probability between each pair of the classes. Furthermore, we will present a ranking algorithm of vertex importance (including active influence and driven influence), which considers the influences of the vertex itself and the indirect influence from its neighbors, and then the key nodes can be identified. The algorithm can be verified through the way of defect planting and infected vertex statistics. Finally, for the purpose of providing a quality assessment method for complex software systems, the Fault-tolerance will be quantitative analyzed through the empirical studies on defect planting, the Fault-tolerant Assessment and Risk Assessment will be presented in this work.

目前软件系统的规模越来越大且复杂性越来越高，评估软件系统中类的影响力并识别关键个体，进而对其进行重点监控，以及量化评价软件系统质量，对降低软件开发维护的风险和提高软件质量具有积极意义。本课题拟以开源软件系统为研究实例，根据类间方法调用关系将源代码抽取为加权软件网络，对软件网络中的关键节点进行识别和分析。首先根据类间方法调用得到节点（类的潜在缺陷）的传播范围和被传播范围，评估节点的传播效应和被传播效应；综合类自身潜在缺陷的影响力和来自邻居类的间接影响力，提出一种包括自身和外面两方面因素的节点影响力（包括施动影响力和受动影响力）排序方法，并识别关键节点，然后拟通过对真实软件系统进行缺陷植入并统计传播节点数目来验证排序方法的有效性；在此基础上，根据缺陷植入量化分析软件系统容错能力，并设计软件系统容错度和风险度度量指标，为复杂软件系统的质量评估提供一种可行方法。

目前软件系统的规模越来越大且复杂性越来越高，评估软件系统中类的影响力并识别关键个体，进而对其进行重点监控，以及量化评价软件系统质量，对降低软件开发维护的风险和提高软件质量具有积极意义。本课题拟以开源软件系统为研究实例，根据类间方法调用关系将源代码抽取为加权软件网络，对软件网络中的关键节点进行识别和分析。首先根据类间方法调用得到节点（类的潜在缺陷）的传播范围和被传播范围，评估节点的传播效应和被传播效应；综合类自身潜在缺陷的影响力和来自邻居类的间接影响力，提出一种包括自身和外面两方面因素的节点影响力（包括施动影响力和受动影响力）排序方法，并识别关键节点，然后拟通过对真实软件系统进行缺陷植入并统计传播节点数目来验证排序方法的有效性；在此基础上，根据缺陷植入量化分析软件系统容错能力，并设计软件系统容错度和风险度度量指标，为复杂软件系统的质量评估提供一种可行方法。