Abstract
Cloud computing has become an important scientific computing and commercial application paradigm. Many computing resources and data exist in clouds, and face various trust issues. Malicious providers may provide poor services to users, while malicious users may give good providers unfaithful trust evaluations. Hence, it is important to detect these malicious nodes that can be providers and users. Current studies on trust management do not sufficiently address the issues of minimizing trust management overhead and maximizing the ability to detect malicious nodes. This paper proposes a new trust model and related algorithm to decrease trust management overhead and improve malicious node detection ability based on domain partition. Partitioning nodes into domains is helpful for decreasing the overhead of trust management in terms of trust storage and computation. Domain and cross-domain sliding-windows are proposed and utilized to store the most recent trust values. Then, an algorithm is designed to compute domain and cross-domain trust values for nodes, and a filter procedure is adopted to remove malicious trust evaluations and malicious nodes from a domain. Simulation results show that the proposed model and algorithm outperform two updated methods, i.e., one based on fuzzy mathematics and another based on authenticated trust and reputation calculation and management, in terms of speed and accuracy of trust computation and malicious node detection.