The Research On The Ciphertext Retrieval Technology Based On Search History

With the rapid development of cloud computing, cloud storage has been widely used by more and more areas. Through effective integration of storage resourses, cloud storage can provide flexible storage service and unlimited storage resources. In cloud storage, the data is mananged by the cloud service provider, and the cloud service provider can get the data easily without access control. Because the server is untrusted, the problem of data security is getting worse. The user's data must be stored in the form of ciphertext in the cloud.When the ciphertexts grow up to a big scale, we need to solve another problem: how to retrieve the ciphertext. The user wants to retrieve the ciphertext quickly.Therefore, the research on the ciphertext retrieval is very important.In this paper, we introduce the cryptography, ciphertext retrieval and the technology of sorting search results first. We analyze the shortcomings of the ciphertext retrieval in the actual application. And we optimize relevant technologies of ciphertext retrieval to inprove the performance finally.We introduce the ciphertext index building technology first, and then we propose an improved ciphertext retrieval strategy. In index update, the geometric partition merge is not good enough for deleting documents. So we improve the geometric partition merge and propose a new geometric partition merge based on bit array. The optimized algorithm can effectively support for deleting documents, and improve the performance of the index.The ciphertext retrieval technology is lack of effective mechanism to sort the search results now. And the search results are mixed with a large number of irrelevant ones. For this deficiency, we propose the ciphertext retrieval technology based on search history. We set up models for the ciphertext and user's interest, and generate the ciphertext feature vector and user interest vector for the ciphertext and user. We measure the similarity between the ciphertext and the user by calculating the correlation between vectors, and optimize the ciphertext retrieval results by the correlation value. We update the user's interest model by recording the user's search history during the retrieval process.Finally, the ciphertext retrieval prototype system has been designed and implemented based on the optimization techniques above. Compare the testing results, it could be found that the retrieval efficiency of the index file and the precision ratio of the ciphertext retrieval have been greatly improved after the optimization.