网络化软件交互行为动态建模

摘要

Current behavior modeling methods have its limitation, to deal with more complicated and uncontrollable networked software behavior, model definition and optimization need to be set up. The invariant constraints mining algorithm was investigated, six types of invariants are mined from the software interactive behavior log, which simplifies the model space. To improve the data set expression capacity, events was adopted to describe the states, and data value relationship was mapped into control flow, which more realistically describes the interactive behavior,meanwhile,events in the model satisfied with invariant constraint rules, which provide the basis for software behavior analysis. To ensure the certainty and compleueness of the model, the method of merging equivalent states in the divided sub-diagram was proposed. The experimental and analysis results show that the effectiveness and feasibility of the methods are validated.%目前的软件行为建模方法有其局限性,而网络化软件交互行为比传统软件更为复杂难控,对模型的定义和优化提出了更高的要求.本文提出一种基于不变量约束规则的挖掘方法,从监控收集的软件交互行为日志中挖掘出六类不变模式,简化了模型空间；并用事件描述状态,提高了数据集的表达能力；构建的动态模型考虑了参数传递和组件之间的交互,将数值关联关系映射到控制流中,从而更真实地刻画了软件交互行为；模型中的事件满足不变量约束规则,为软件行为分析提供了依据；同时,本文提出合并划分子图间的等价状态方法,对模型进行精化和抽象,确保了模型的确定性和完备性.仿真实验和实例分析证实了该方法的正确性和有效性.