Secure Sketch Search for Document Similarity

Document similarity search is an important problem that has many applications especially in outsourced data. With the wide spread of cloud computing, users tend to outsource their data to remote servers which are not necessarily trusted. This leads to the problem of protecting the privacy of sensitive data. We design and implement two secure similarity search schemes for textual documents utilizing locality sensitive hashing techniques for cosine similarity. While the first one provides very fast search time results and a decent level of privacy, the second method enjoys enhanced security properties such as hiding the search and access patterns but with higher latency.