【作者】 文杏梓；

【导师】 罗新星；

【作者基本信息】 中南大学 ， 管理科学与工程， 2013， 博士

【摘要】 摘要：计算机的广泛应用及网络技术的飞速发展,软件已经渗透到国民经济乃至国防建设的各个领域,成为影响人们生产、生活的重要组成部分。随着软件规模的日益增大,复杂程度越来越高,软件的风险也变得越来越难以控制,各种由于软件缺陷或故障所引起的事件频繁发生。发展高可信软件,已经成为当前国际软件技术发展的战略制高点,引起了全社会的普遍关注。基于多维质量属性的可信软件评价是实现软件可信、开展可信软件管理的核心基础,也是可信软件开发管理过程中急需解决的问题之一。目前,绝大部分软件质量属性的研究都是基于软件设计、开发者的视角,而忽视了软件用户在使用过程中的客观实践及主观感受。因此,研究基于用户需求的可信软件质量属性评价方法具有重要的理论价值和现实意义。本文首先探讨了可信软件、质量属性的界定,及其国内外研究现状,并对几个主要的软件质量模型进行了归纳总结；从软件质量度量的发展、分类及度量步骤三个方面对软件质量度量进行了研究,这些都是可信软件质量属性研究的基础性工作。可信软件质量属性指标体系的产生是软件质量属性评价的基础,也是评价过程中关键性的第一步。针对用户对软件质量属性的需求及可信软件的特性,本文将可信软件质量属性分成关键属性与非关键属性两大类。在此基础上,通过构建可信软件质量属性证据模型、评估体系,形成满足用户需求的可信软件质量属性评价指标体系。这是可信软件质量属性评价方法的基础。本文提出的可信软件质量属性评价的第一个方法是一种间接的、考虑软件使用者和软件设计开发者一致性评判的研究方法。针对可信软件的某些质量属性不可直接测量、质量属性之间不具备可比性且测量标准不统一等问题。本文研究了质量属性之间的相互关系,通过设计结构矩阵、矩阵转换、矩阵运算来解决上述问题并间接度量软件质量属性。同时,由于可信软件运行环境的不同,不同的人对于质量属性有着不同的视角、解释和判断标准,在很大程度上人们没有办法达成一致的评判准则。针对这个问题,本文通过上述间接度量模型及三个评价指标,来确定软件设计开发者和软件用户对于软件质量属性评判的一致性程度,并根据设计的准则,做出相应的决策。与其他方法相比,这种方法完全依赖于软件质量属性之间的相互关系,是一种客观的、具有统一标准和评价尺度的、考虑软件使用者和开发者一致性评价的方法。可信软件质量属性评价的第二个方法是一种直接的、考虑评价者有限理性的研究方法。现今绝大部分评价方法所假设的一个基本前提是评价者是完全理性的。但实际上评价者在评价过程中带有主观偏好及风险性是不可避免的。本文引入前景理论以描述、表达评价者的有限理性。通过在“框架”内,正负理想参考方案的选择、信息的处理、综合前景值的确定、比较,来判断软件质量属性的优劣。同时,考虑在评估过程中,评价者不同偏好对评价结果的影响,本文对其作了进一步的分析研究,从而使得评估方法更加严谨、科学。可信软件质量属性评价的第三个方法是一种动态的、考虑软件在受到干扰的环境下,其质量属性评价值是否发生变化、如何变化、软件是否总是可信的评价方法。软件系统是一个动态的系统,它总是在不断的发展、演变中,在受到干扰时,我们总是期望它能保持原来的状态,但实际上这并非总是可能的。系统运行状态的变化,必然导致系统内部能量的变化,本文创新地借助于物理学“熵”的概念来描述、表达这个现象,并针对软件状态的变化是一个时间、空间完全离散的,且后一个状态是受前一个状态影响的特征,结合元胞自动机的相关机理,对软件质量属性在受到干扰后的评价值进行了有效模拟。本文所提出的方法,均是基于软件用户需求的视角,对可信软件质量属性进行间接、直接、动态的全方位评价,以期有利于软件用户对可信软件质量属性进行客观评价,也有利于软件设计者在设计过程中更多的考虑用户需求提供客观依据。图54幅,表14个,参考文献更多还原

【Abstract】 Abstract:With the widespread use of computer and rapid development of network technology, software has penetrated into all fields of the national economy as well as national defense construction, and been a very important part of human life. Followed by the increasing the size and complexity of software, it is very difficult to control the risk of software and is very frequent that the disastrous events have happened due to software defects or failures. Development of High Confidence Software has become a strategic high ground of the current international software technology development, which has caused a widespread concern in the human society.Trustworthy software evaluation based on multi-quality-attributes becomes the core foundation to achieve software trustworthiness and carry out the management of trustworthy software, which is also an urgently resolved problem in the process of trustworthy software development and management. Currently, the major research on quality attributes of software is based on the developer’s perspective, ignoring the objective practice and subjective experience of software users in the process of using software. So the research based on user requirements in the field of trustworthy software has become very important for theory and practice.This paper discusses the definition of trustworthy software and quality attributes firstly, particularly pays attention to overseas and domestic research status about them, and summarizes several major software quality models, researches on software quality metrics from three aspects of development, classification and procedure. All these are the basic work to research trustworthy software quality attributes.It is not only a basic work but also a crucial step to build the index system of trustworthy software quality attributes for evaluating them. Based on the users’requirements and the features of trustworthy software, this paper divides the quality attributes into two categories, which are key attributes and non-key attributes respectively. And after building the quality attributes evidence model and the quality evaluation system for trustworthy software, an evaluational criteria system for meeting user requirement is built. All these are the basis to evaluation method for trustworthy software quality attributes.The first evaluation method of trustworthy software quality attributes in this paper is an indirect method, which considers that software users and software developers should have a consistent evaluation for quality attributes. Because some quality attributes can’t be measured accurately or be compared directly, this paper researches the interrelation of quality attributes in the component, proposes design structure matrix to describe their interrelation, and computer their relative importance by matrix shift. Meanwhile, the running environment of trustworthy software is different, and different people have different perspectives, different explanation and different criteria, they can’t gain the consistent criteria to a large degree when evaluating quality attributes. With respond to these difficulties, this paper poses an evaluation method based on above measurement model and three indices. Compared with other methods, our approach depends on the interrelation of quality attributes, which make result objective and have a unified standards and evaluation criteria, considering the consistency of software users and developers for evaluation result.The second evaluation method of trustworthy software quality attributes in this paper is a direct method, which considers the valuators’ bounded rationality. Most evaluation methods in the past have suppoosed that the evalutor are entire rationality, but in fact all kinds of subjective factors, such as risk preference, subjective preference, affect the valuator’s evaluation result. This paper introduces the prospect theory, which is widely used in the field of psychology and behavioral science. Found the positive, negative ideal solution in the framework, the comprehensive prospect value of software quality attributes can be calculated and ranked to find the best software quality attributes. Because in the evaluation process the risk preference from valuators can affect result, we further analyze relevant parameters in the prospect theory, which make our method more rigorous and more scientific.The third evaluation method of trustworthy software quality attributes in this paper is a dynamic method, considering whether the evaluation result of quality attributes will change, how change, and whether software is trustworthy when the software suffers interference or continuous interference. Software system is a dynamic system with constant development and evolution, when disturbed, the system always is expected to maintain the original state, but in reality it is not always possible. It is inevitable that energy changes in the system with the changing of statement, the paper uses entropy, a physics concepts, to describe the phenomenon. When software suffered interference, its status change shows that it is completely discrete process in time and space, and the subsequent state is affected by previous one. The cellular automation can meet these conditions and simulate this process to gain the evaluation result of software quality attributes.This paper evaluates the trustworthy software quality attributes based on the perspective of software user requirement indirectly, directly and dynamic. It is benefit to evaluate the quality attributes objectively for the software users and to provide an objective basis for software developers, who consider users’requirement a bit more in the process of designing the software.更多还原