The Research Of Inference Control Algorithm For RDF(S) In Semantic Web

Semantic Web is an intelligent network which is able to make judgement and inference based on semantics. People may reason out unauthorized data from those that have already authorized. Therefore, further study of the Semantic Web inference control algorithm has application value and is important. There are still some deficiencies of operational efficiency and safety that need to be further improved and enhanced in the inference control algorithm.In this thesis, in order to improve the efficiency of inference control algorithm, we firstly propose a method of tidying RDF data source by sorting prior to inference control; secondly, we analyze and point out that RDF(S) triples inference control algorithm proposed by Lu et al. can affect the user access to some insensitive information normally while preventing illegal reasoning effectively, and on this basis,we propose two improved algorithms. Details are as follows:1. We tidy RDF data sources by sorting according to the feature of RDF graph, and when the data source update, the new data can be classified quickly. In order to avoid the unnessary time comsumption, the system should firstly determine whether the current triplet information needs the inference control process.2. We propose two improved RDF(S) inference control schemes to make the insensitive information completely shielded in reasoning process as little as possible. In the first scheme, the inference conditions of the sensitive information are marked, and those marked triples can only be accessed instead of being used for inference. In the second scheme, the inference rules are forced to change using Jena framework and the inference results of the sensitive information are pointed to the inference conditions.3. We accomplish the improved algorithm using the Jena framework. The results show that:taking tap-cmu.rdf and tap-cmu-scheme.rdf data source (736pieces of semantic information) as examples, thousands of millisecond-level time is reduced; for different collections of sensitive information, the improved algorithm can retain3-15pieces of masked semantic information in the original inference control algorithm.