Trust and reputation systems (TRSs) are used in cooperative environments where an agent needs to make a decision for requesting or performing a service. However, TRSs can be abused by malicious agents who do sequences of dishonest actions (attacks). Although there are proposals on verification of TRSs against attacks, they are not comprehensive enough to evaluate various Trust Computation Models (TCMs) and/or do not provide sufficient expressive power to specify different required robustness properties. In this paper, we introduce a comprehensive framework for specifying and verifying various robustness properties in TRSs through model checking approach. The proposed framework includes three main parts: (1) a logic for specification of robustness properties of TRSs named Probabilistic Action and Reward based Computation Tree Logic (PARCTL), (2) an enhanced version of our previously proposed model for specifying TRSs in hostile environments named Trust and Reputation Interaction Model (TRIM), and (3) the required algorithms for quantitative and probabilistic model checking of PARCTL properties over the specified model. The proposed framework has been implemented as a tool named TRIM-Checker. Our experimental results on robustness analysis of famous eBay, Beta, and CORE TCMs using TRIM-Checker are presented and their robustness against attacks is evaluated and compared together.

信任和信誉系统（TRS）用于合作伙伴需要代理做出请求或执行服务的决策的环境中。但是，TRS可能会被恶意代理滥用，这些恶意代理会执行一系列不诚实的操作（攻击）。尽管有针对TRS进行攻击验证的建议，但它们不够全面，无法评估各种信任计算模型（TCM），并且/或者没有提供足够的表达能力来指定不同的所需鲁棒性属性。在本文中，我们介绍了一个全面的框架，用于通过模型检查方法指定和验证TRS中的各种鲁棒性。提议的框架包括三个主要部分：（1）一种用于指定TRS鲁棒性的逻辑，称为概率动作和基于奖励的计算树逻辑（PARCTL），（2）我们先前提出的模型的增强版本，用于在称为信任和信誉交互模型（TRIM）的敌对环境中指定TRS，以及（3）在指定模型上对PARCTL属性进行定量和概率模型检查所需的算法。提议的框架已作为名为TRIM-Checker的工具实现。给出了我们使用TRIM-Checker对著名的eBay，Beta和CORE TCM进行鲁棒性分析的实验结果，并评估并比较了它们对攻击的鲁棒性。提议的框架已作为名为TRIM-Checker的工具实现。给出了我们使用TRIM-Checker对著名的eBay，Beta和CORE TCM进行鲁棒性分析的实验结果，并评估并比较了它们对攻击的鲁棒性。提议的框架已作为名为TRIM-Checker的工具实现。给出了我们使用TRIM-Checker对著名的eBay，Beta和CORE TCM进行鲁棒性分析的实验结果，并评估并比较了它们对攻击的鲁棒性。