VLMC控制流软件可靠性度量方法

针对现有软件可靠性模型普遍不适用于实际软件的问题，分析了软件内部模块间控制转移机理，通过分离入口模块，提取一阶控制转移概率矩阵；弱化内部模块与出口模块间差异，重构二阶以上转移概率矩阵，在改进Cheung模型的基础上建立一个更符合软件实际的可靠性度量模型。针对变阶依赖（ VLMC）控制转移导致的可靠性度量难问题，通过对导致复杂依赖的多入多出模块进行节点扩展，将VLMC控制流转化为Markov链，利用所建立的软件可靠性模型对VLMC控制流软件进行可靠性度量。研究利用演绎推理对所建立模型进行了正确性形式化证明。最后给出了方法的实例验证。

Reliability Measurement for Software with Variable Length Markov Control Transfer Chain

The classical software reliability model of Cheung based on architecture analysis is imperfect when applied to software which has non-uniqueness terminal node because of ideal assumptions. In order to handle this problem, an improved model is established through two improvement as follows: first, probability matrix of control transfer from start node to other inner nodes in a given software is computed independently of control transfer within inner nodes on the assumptions that the start node is inaccessible to control transfer from any inner nodes; and then, probability matrixes of control transfer within inner nodes are reconfigured by weaken the difference between inner nodes and ending nodes. In order to measure the reliability of software with VLMC control transfer flow, the variable length Markov control sequence chain is converted to simple Markov chain by node expansion and conditional transfer probability redistribution. By using deductive inference, a formal proof for the new model is given in this essay. In the end, effectiveness and simplicity of the proposed method is verified by a software example.