基于改进投影寻踪技术的软件可信测度模型研究

发布时间：2018-11-17 10:24

【摘要】：软件已经应用到政治、经济、军事、文化领域等方方面面。由于软件规模、复杂度和智能化水平不断提高,应用环境日益复杂,以及软件开发过程的不可控性,软件产品上线运行存在着诸多缺陷,导致软件不可信。软件可信测度方法是近年可信软件领域研究的重点和难点问题之一。通过文献梳理和软件企业调研,系统地研究了国内外可信软件测度相关最新研究成果,软件可信属性主要侧重于保密安全性、生存性、容错性、可靠性、可靠安全性、实时性等方面研究。由于软件可信属性的多样性,且属性间具有不一致性,单一属性可信很难全面反映软件整体可信水平。为了进一步探讨软件可信属性之间的关系,通过对软件可信属性的多样性和软件可信测度模型的复杂性分析,根据ISO/IEC25010软件质量标准,建立基于软件行为可信、能力可信、用户行为可信的软件可信指标体系,充分考虑不同软件对可信测度要求侧重点不同,可组合调整可信测度指标项。由于可信软件测度指标属性具有高维特征,针对投影寻踪方法中最佳投影方向优化问题,引入引力搜索算法(Gravitation Research Algorithm,GSA)的思想,采用罚函数法(SUMT)构造新的投影指标函数,简化了投影指标函数的复杂程度,结合层次分析法和熵权法,将高维可信软件属性特征降维成有效的低维数据,提出了基于改进投影寻踪技术的可信软件测度模型。为了更加直观的反映可信软件等级,借鉴近邻聚类传播算法,提出了一个可信软件分级方法。案例研究表明,综合软件可信测度模型运算结果与软件实际运行情况基本相符,为有效测度大型复杂软件系统可信水平提供了一种新的方法。
[Abstract]:Software has been applied to political, economic, military, cultural and other areas. Due to the continuous improvement of software scale, complexity and intelligence level, the increasingly complex application environment, and the uncontrollability of software development process, there are many defects in the on-line operation of software products, which lead to the software not credible. The method of software confidence measurement is one of the key and difficult problems in the field of trusted software in recent years. Through literature review and software enterprise investigation, this paper systematically studies the latest research results of trusted software measurement at home and abroad. The trusted attributes of software mainly focus on security, survivability, fault tolerance, reliability and security. Real-time research. Because of the diversity of the trusted attributes of software and the inconsistency among them, it is difficult to reflect the overall level of trust in software by single attribute trust. In order to further explore the relationship between software trusted attributes, the diversity of software trusted attributes and the complexity of software trusted measure model are analyzed. According to the software quality standard of ISO/IEC25010, the software behavior is trusted and the ability is trusted. The trust index system of software based on user behavior trust, considering that different software needs different trust measure, can combine and adjust the confidence measure index. Because the attribute of trusted software measure index is of high dimension, aiming at the optimal projection direction optimization problem in projection pursuit method, the idea of gravitational search algorithm (Gravitation Research Algorithm,GSA) is introduced. The penalty function method (SUMT) is used to construct a new projection index function, which simplifies the complexity of the projection index function. Combining with the analytic hierarchy process (AHP) and entropy weight method, the attribute features of high-dimensional trusted software are reduced to effective low-dimensional data. A trusted software measurement model based on improved projection pursuit technique is proposed. In order to reflect the level of trusted software more intuitively, a classification method of trusted software is proposed based on the nearest neighbor clustering propagation algorithm. The case study shows that the calculation results of the integrated software confidence measure model are basically consistent with the actual operation of the software, which provides a new method for effectively measuring the trusted level of large-scale complex software systems.
【学位授予单位】：山东工商学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.5

【相似文献】