Research On Trust Management Mechanism And Its Application In Distributed Open Environment

With the rapid development of network technology, the need of resource share and interactive collaboration from all kinds of users in distributed open environment is increasing dramatically. Although the work is more efficient, the security problems bought by the openness, dynamics, complexity and collaboration of the environment should not be ignored. On one hand, dynamic trust relationship should be established between two parts in an interaction, to protect the shared resource from malicious destroy. On the other hand, based on the trust relationship, trusted decision about whether or not to provide resource service should be made, to establish effective interactive collaboration environment. Faced with the new requirements, the drawbacks of traditional security mechanisms for closed and central environment are exposed. Trust management mechanism, which helps participants without enough prior knowledge to establish healthy and effective collaborative relationship, has become one of the core supporting technologies to resolve the security problems in distributed open environment.The trust relationship between interactive entities in distributed open environment is subjective and uncertain. Based on the cloud model theory, an uncertainty enhanced subjective trust evolution strategy is proposed. Firstly, a series of definition, such as trust value space, trust concept space, trust cloud, trust concept cloud, sincerity value space, sincerity concept space and sincerity concept cloud, are defined. Then, three basic subjective trust operators, i.e., forward cloud operator, backward cloud operator and similar cloud operator, are designed. Subsequently, the algorithms to realize trust propagation, trust aggregation, trust assessment and trust update respectively are proposed. The time complexity and correctness of the four algorithms are analyzed theoretically. Lastly, the effectiveness, rationality and efficiency of proposed strategy are verified on Repast simulator. The results show that the similar cloud operator and similarity degree used by the trust evaluation algorithm evaluate the trust level effectively; Behavior characteristics of different entities considered are reflected accurately, and malicious entities are chosen gradually by the trust update algorithm. Compared with the other mechanisms possessing similar functions, the proposed strategy supports more perfect functions on trust evolution without much loss of efficiency, provides more chances for entities to safely cooperate with the others, and captures and reflects the characteristics of entities'behavior more sensitively and accurately.In distributed open environment, whether or not the resource owner provides the resource requestor with resource service is decided not only by the trust relationship of two parts, but also by the risk existing in the execution of service process. In a word, both trust and risk affect the decision making of resource owner. The distinction and relation between trust and risk are emphasized. A method to quantify the risk is proposed and exploited for the security decision in resource service. Firstly, the pivotal source to evoke risk in trust management is discussed, the relationship between trust and risk is analyzed, and the drawbacks of the existing ways to quantify risk are indicated. Then, based on a group of formal definitions associated with trust and risk, a model for risk-aware resource service decision making is proposed. Subsequently, a reputation inference process based on probability theory is put forward, and a risk degree algorithm based on Chernoff Bound theorem is designed. The method is employed to assess the trustworthiness of resource requestor, and the decision on whether or not the resource can be accessed by the requestor is made according to the evaluation result. The time complexity and correctness of the risk degree algorithm are analyzed theoretically. Extensive experiments are conducted on Repast simulator. The simulation results show that the proposed method estimates the trustworthiness of resource requestor rapidly and accurately, and assists the resource owner in trusted decision making for resource service.In order to validate the effectiveness of the proposed subjective trust management mechanism, an emerging distributed open environment-data-intensive computing environment is chosen. Taking a kind of typical data-intensive application-Bag of Tasks as an example, a reliability-aware multiple criteria replica selection strategy is proposed. The strategy does an overall consideration of the reliability, usage cost and fetching time of datasets required by task execution. The reliability is quantified by the trust evolution mechanism based on the cloud model theory. The construction of the set of best data replicas is abstracted as a certain multiple criteria decision making problem. Technique for Order Preference by Similarity to an Ideal Solution is used to resolve above problem. The time complexity and correctness of the reliability-aware replica selection algorithm are analyzed. Extensive experiments are conducted on OptorSim simulator. The simulation results show that in an unreliable environment, the replicas with higher reliability are chosen, thus the losses of time and economic cost caused by the failure during accessing data are decreased; In a reliable environment, the performance of the proposed strategy is similar to the perfect condition, which means that the strategy is flexible to the change of the environment.