Research On Reputation System Supporting Privacy Preserving And Security Data Processing

The high-speed development of the Internet has brought high-quality network services to the daily life of human beings.Reputation system provides a reputation mechanism for each node in the network,satisfies the evaluation of the information quality through a clear reputation value among the nodes.The reputation system also stimulates the civilized behavior of the nodes,and can make some predictions about the future behavior of service providers and service consumers at the same time.The reputation system provides people with excellent network services,so it has been widely used in network environments such as resource sharing,e-commerce and public forums etc.However,in most practical applications,the evaluating information provided by nodes also reveals many of the evaluator's private information.With the advent of the age of data,if an attacker obtains these information and uses it maliciously,it will be detrimental to the evaluator and may also break the entire system in serious cases.In the reputation system,the correlation between the node's identity and the evaluating information needs to be lowered to achieve the purpose of protecting the privacy of the node;but in order to ensure the credibility mechanism of the reputation system and simultaneously monitor the malicious behavior,the evaluating information of the node needs to be associated with the identity.It seems to be a paradox.Therefore,it is necessary to set a reasonable privacy preserving mechanism while the reputation system can provide a good reputation-based network service.As a special network model,the reputation system has attracted widespread attention from scholars at home and abroad.This thesis presents a new type of reputation system supporting privacy preserving and security data processing,based on the research results of the trust model,the mechanism of reputation scores,and the mechanism of privacy preserving in the reputation system,utilizing a series of cryptographic tools,anonymous network models,and machine learning concepts.The main contributions of this thesis is summarized as follows:(1)This thesis analyzes and improves an existing reputation system model,proposes a new privacy preserving mechanism based on a special anonymous network environment,and subtly applies the technology of secret key sharing and revocable ring signature.Our system satisfies the request of saving a lot of calculation steps,reducing calculation overhead,and improving system performance meanwhile.(2)When malicious behavior occurs,current reputation systems have not achieved an approach of classifying and handling diverse behaviors.We imported the idea and mechanism of machine learning,categorized different types of malicious behavior,and traced different malicious behaviors according to the special mechanism of revocable ring signature.At the same time,we implemented and simulated the system based on Java development environment and Matlab software under Windows system.The results show that our system outperforms the existing system in terms of calculating overhead,and performs high-precision classification procedure on user nodes' behaviors.