Towards Efficient Yet Privacy-Preserving Approximate Search in Cloud Computing

Owing to the great advances in cloud computing and Internet technologies, data owners (DOs) have been motivated to outsource the storage of their data to remote cloud servers (CSs) in order to enjoy great data management service with an efficient cost. For security purposes, DOs usually have to encrypt their data prior to outsourcing it to the untrusted CSs. But encryption makes searching the encrypted data a challenging task. Recently, several approaches have been provided to enable searching over encrypted data. However, the majority of these systems are limited to handling an exact search, not a similarity search; but the latter is an important need for real-world applications. In this paper, we propose an efficient yet secure scheme to search encrypted cloud data, while recovering the misspellings and typographical errors that exist frequently both in the search request and in the source data. To do so, we use a metric space to construct a tree-based index, which allows retrieving only the relevant entries with a minimum number of distance evaluations. String embedding techniques are used to refine the relevant entries efficiently and securely. Our index construction maintains the privacy of the keyword trapdoors as well as the stored data. Comparing our scheme with other similarity searchable encryption systems via experiments shows that our scheme is efficient in terms of search time and storage overhead.