Research On Key Techniques Of Building Component-based Software System With High Trustworthiness

With the continuous development of computer technology,computer software is widely used in various fields.However,due to the complexity of the problem to be solved,changing of user needs,and dynamic evolution of operating environment,increasingly large-scale software systems often occur various failures,software trustworthiness becomes increasingly prominent.In order to reduce the losses caused by the software incredibility,how to improve the trustworthiness of the software becomes a hot topic at home and abroad.For purpose of software trustworthiness researchers made a series of research results on the basic theoretical models,security control,and test and measurement standards of trusted software,but little research show concern on software life cycle to consider software trustworthiness.Based on the above situation,under the guide of software architecture and trusted software,and from the perspective of constructing trusted software,this papaer study some key technologies of building high-confidence software systematicly in top-down software development process.Firstly,we introduce trustworthy attribute formally in ADL to ensure the software trustworthiness at system overall design stage;secondly,we design a trustworthy model of component to ensure the trustworthiness of component at detailed design stage;then we design a trust model in order to select reliable third-party components efficiently when OTS component needed;finally,we design a self-healing model so the system can resume automatically when it fails.This study has made the following innovative achievements:1.Based on current status of little support on modeling trustworthy attributes in ADL,we describe how to model measurable trustworthy attributes formally using TLA+,present a new architecture description language RXF-ADL based on the TLA+ formal semantics,and implement the corresponding supporting tool,thus we can make estimates for the timeliness and reliability of the whole system in overall system design stage.2.We present a component trustworthy model,which describe the various trustworthy properties of the component formally with set theory and predicate logic,and design some conversion rules used to convert component dependability model to extended timing automata,then we can achieve automatic verification of trustworthy attributes using UPPAAL tools in detailed design stage.The dependability model is independent of the existing component development technologies,namely programmers can use any component technology to implement component.3.We propose a component trust model based on trusted graph,which defines simple and efficient trust computing and evolving rules,and then implement a component selection algorithm,so we can choose reliable OTS component effectively in component-based software engineering at implementation phase.4.For the management requirement of component statistics according to trust interval,we propose a new B-tree named mca B-tree,which can multi-class range statistical issues efficiently.We all know that B-tree is an efficient structure to solve range statistical query,its query and update complexity is O(log CN)I/Os,here N is total data amount,C is disk page size.It is evident that Q times range statistical query needs O(Qlog CN)I/Os.However,the mca B-tree can answer multi-category range statistical query in O(log CN)I/Os,and its update performance is also O(log CN)I/Os.5.We propose a self-healing engine based on aspect,which can solve the problem of run-time system component's malfunction.After the integration of self-healing engine,self-healing engine will automatically handle exceptions so that the system returns to normal operation transparently if some component is abnormal.