Determination of Parameters Balancing between Security and Search Performance on Searchable Encryption

In the searchable encryption, searching encrypted data is slow, compared to searching non-encrypted data. As a solution, there is an acceleration method that discloses a small portion of a hash value of keyword to be searched. The disclosure bit(s) allows the cloud to narrow down the search space and reduce the amount of computation. The more bits are disclosed, the faster the search speed becomes. However, repeated disclosure of part of hashed keyword in queries leads to reveal the frequency distribution of plaintext keywords. The longer the disclosure bits are, the more accurate revealed distribution becomes. In this paper, we propose a method to determine the disclosure bit length by using minimum entropy and k-anonymity. Furthermore, scalability of the proposed method is enhanced by employing distributed databases. To show that the proposed method achieves both high acceleration effect and security, we show an evaluation result of the proposed method applied to the conventional full-text search system using searchable encryption. Our evaluation result shows that the proposed method reduces the search time of the system by up to 97.2% when the number of stored document files is 1,000. While a database in this evaluation is designed so that disclosure bits satisfy at least 31-anonymity by using the proposed method, we observe that the database actually satisfies 2,598-anonymity.