Influences And Control Of Distrustable Factor On Software Trustworthiness

With rapid development of information industry, software has been ubiquitously infiltrated into several areas, such as national economic, science and technology development, and national defense construction, so software plays an important and basic role in information society. However, as demands on software functions increase continuously, software has become more and more huge and complex, and defects and loopholes are hard to be effectively controlled. In the open changing running environment, software system becomes extremely fragile, and various faults and failures even disasters arise frequently, which cause different losses for users. Therefore, software is not always trustable, both academic and industry from domestic and abroad show high concerns with software trustworthiness.Both software development and software system itself are complex systems. Although development theory and technique have been made great progress, the accident resulting from software trustworthiness never ceases. Hence, the problem of why software is untrustworthy should be further analyzed. Based on the basic concept of distrustable factor which is defined as the original cause of software distrust, and some academic methods that are Differential Dynamics Theory, Systematic Dynamics Theory, Expected Utility Theory, Queuing Theory and Petri Net Modeling Principles etc., this dissertation focuses the study on the models of action mechanism, effect and control of distrustable factor during software development process, the main contents are as follows:1. Action mechanism of distrustable factor. Based on the analysis on implications of some concepts relevant to "distrust", the action mechanism of distrustable factor which follows a distrust chain is proposed. The chain of "distrustable factor—error—defect (fault)—failure—distrust" is called as distrust chain, which extends the classical failure chain from software reliability theory. Meanwhile, with some given datasets, the correlations of distrustable factor with distrustable factor, defect type and trust principle are presented. The results help to understand the distrustable factor clearly and enrich its concept system.2. Prediction of software defect based on distrustable factor. As one kind of defect is concerned, considering the defect growing process, a model of defect states (i.e., be latent, be founded, and be repaired) transferring is constructed. The latent defect is estimated from the stable equilibrium solutions of the model, and an instance is taken to exemplify the effectiveness of the model. As the whole defect is concerned, considering software reliability growth, an improved G-O NHPP model is established with the assumption that acting strength of distrustable factor on defect follows a nonnegative distribution. Compared with other classical NHPP models, some examples illuminate that the proposed model has more fitting and predictive power.3. Impact of distrustable factor on software process product quality. Software quality is only guaranteed during software development process, and final product quality depends on process product quality. Taking the waterfall development process for instance, a systematic dynamic model for analyzing the impact of distrustable factor is presented. According to the simulation of the model, the range and degree of impact from various distrustable factors on process product quality are observed and measured.4. Evolution of software trustworthiness affected by distrustable factor. Firstly, one type of fault is synthetically measured by its severity, duration and frequency of occurrence. Secondly, with expected utility theory, a new method for measuring software trustworthiness based on fault utility is proposed. Borrowed insights from queuing theory, a model is given to measure software trustworthiness in the duration of one kind of fault. Finally, for different occurrences of two kinds of faults caused by distrustable factor, the dynamic evolution of software trustworthiness is presented. A SIS developed by some software industry is taken for instance, the evolution of system trustworthiness during its test running phase is analyzed.5. Control process of distrustable factor based on Petri net. Concerning the pre-control of distrustable factor, the factor of "experience and skill level of developer" is chosen, and a colored stochastic Petri net model is constructed. In order to control the probability of making mistake from developer, the model can simulate the effect of personnel training of software industry. Concerning the ex-post control of distrustable factor, the factor of "requirement changing" is chosen, and to control the average delay time of requirement changing, a stochastic Petri net model with fuzzy time parameter is established, and the time performance analysis results of the model are presented.There are some researches indicate that many failures of software projects is due to the problem of management, rather than technology. With the principle of management, some researches on software trustworthiness based on distrustable factor are performed in this dissertation. The findings can offer some ideas on studying software trustworthiness from a new view of software distrust, a new method for measuring and evaluating software trustworthiness, and an effective way for dynamic monitoring software trustworthiness. Meanwhile, the findings can supply some references for scientific management and continuous process improvement of software project.