基于属性评测的软件可信性度量技术研究

发布时间：2018-10-18 15:11

【摘要】：随着计算机科学在政治、经济以及科技等领域的广泛应用,软件技术已应用到了人们生活的方方面面。软件变得更大更复杂,软件质量也面临着更大的考验,对软件可信水平评价的软件可信性度量技术也变得越来越重要。然而目前软件可信性度量技术并不成熟,其表现为软件可信性没有统一定义、软件可信属性没有统一标准、没有成熟的软件可信属性度量模型,并且现有的软件可信性度量方法也存在着理论难以实践以及度量结果不精确等问题。针对上述问题,论文提出了一种基于贝叶斯网络的软件可信性度量方法,该方法利用贝叶斯网络在体系结构上对软件强大的解释和推理能力,根据软件的体系结构建立贝叶斯网络,从结构上将软件分治处理,对贝叶斯网络中的软件组成模块进行度量,得出被度量模块的相关可信指标的参数分布。然后,在此可信指标基础上,根据软件集成模块之间的依赖关系建立贝叶斯网络,获得集成系统的可信指标结果分布。论文主要工作如下:首先,论文分析了可信属性之间的关系,为综合软件各可信属性对软件进行度量奠定基础。其次,为确保可信软件度量的操作可行性,论文提出了一种基于软件体系结构建造贝叶斯网络的方法。论文研究了软件模块化表达方式,包括对软件的类、模块之间的继承、调用等关系的分析,根据它们之间的依赖关系建立了贝叶斯网络,将软件的度量由系统级规约到模块级,减小度量的难度。再次,基于所建立的贝叶斯网络对软件各组件进行度量。论文验证了软件各功能点,以得到功能属性和验证软件集成水平。在可靠性方面度量了软件的成功路径指数和执行任务成功指数。为较快得到准确的度量结果,度量时使用了先验概率方法。对软件运行有重大影响但出现频率低,度量难以覆盖的情况,采用重要性取样原理对其进行加速度量,以得出全面的度量结果。在软件可维护性属性度量方面评价了可维护子属性,探索了可维护性在贝叶斯网络中的度量应用。最后,对航行器多协同路径规划仿真软件进行了可信属性度量实验,使用贝叶斯网络将被度量的软件分治,降低了度量难度。采用贝叶斯网络针对性的度量软件各部分能降低可信度量时注入的用例数量,实验证明了该方法的可行性和高效性。
[Abstract]:With the wide application of computer science in politics, economy, science and technology, software technology has been applied to every aspect of people's life. Software becomes bigger and more complex, software quality is also facing a greater test, software credibility evaluation of the level of software credibility measurement technology has become more and more important. However, at present, the software credibility measurement technology is not mature, which shows that there is no unified definition of software credibility, no unified standard for software trusted attributes, no mature software trusted attribute measurement model. And the existing software credibility measurement methods also have some problems, such as the theory is difficult to practice and the measurement results are not accurate. Aiming at the above problems, this paper proposes a software credibility measurement method based on Bayesian network, which makes use of Bayesian network's powerful interpretation and reasoning ability to software in architecture. According to the software architecture, the Bayesian network is established, the software is divided and treated from the structure, the software component module in the Bayesian network is measured, and the parameter distribution of the relevant trusted index of the measured module is obtained. Then, on the basis of the trusted index, the Bayesian network is established according to the dependence relationship between the software integration modules, and the distribution of the trusted index results is obtained. The main work of this paper is as follows: firstly, this paper analyzes the relationship between trusted attributes, which lays the foundation for synthesizing the trusted attributes of software to measure the software. Secondly, in order to ensure the operational feasibility of trusted software metrics, a method of building Bayesian networks based on software architecture is proposed in this paper. This paper studies the software modularization expression, including the analysis of the software classes, the inheritance between the modules, the invocation and so on. According to their dependencies, the Bayesian network is established, and the software metrics are reduced from the system level to the module level. Reduce the difficulty of measurement. Thirdly, the software components are measured based on the established Bayesian network. The function points of the software are verified in order to obtain the functional attributes and verify the level of software integration. The software success path index and task execution success index are measured in terms of reliability. In order to get accurate measurement results quickly, a priori probability method is used in measurement. It has a great influence on the software operation, but the frequency is low, and the measurement is difficult to cover, so the importance sampling principle is used to measure the acceleration of the software, so as to obtain the overall measurement results. In the aspect of software maintainability attribute measurement, the maintainability subattribute is evaluated, and the application of maintainability in Bayesian network is explored. Finally, a trusted attribute measurement experiment is carried out on the simulation software of multi-cooperative path planning for aircraft, and Bayesian network is used to divide and conquer the measured software, which reduces the difficulty of measurement. Using Bayesian network to measure each part of the software can reduce the number of use cases injected when the confidence level is reduced. The experimental results show that the method is feasible and efficient.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.5

【参考文献】