Distributed Multimedia Information Retrieval Manner Based on the Statistic Information with Privacy

摘要

We propose and clarify the manner in which a distributed information retrieval (IR) is conducted using a Peer-to-Peer (P2P) technology that has approximately the same accuracy as a current index-integrated search engine with the statistical information calculated from the contents We need to solve the following problems to develop distributed IR services for distributed objects in peers without indices centralized servers. The first issue is efficient IR strategies that are highly accurate services by using the statistic information for indices using the collaboration among distributed peers (nodes). The second issue is the prevent strategy from flow of object information icluded personal information and shared information in groups to unknown peers as the result of IR service. We discuss the generating the statistics information and the scalability of service by using simulations and emulations. The dynamic privacy derivation manner of objects is generated based on the information entropy from catchwords. Our proposed stategy is required and suitable for P2P based peer services.