A cross-lingual secure semantic searching scheme with semantic analysis on ciphertext

Secure semantic searching can search arbitrary query words semantically related to keywords over the encrypted data. However, most of the current schemes perform the matching after query semantic expansion on the plaintext. The existing works rarely support cross-lingual searching that breaks the language barrier to allow a query issued in a lingual to retrieve the documents written in other languages. Meanwhile, most of the previous schemes are difficult to be applied to real-time applications because of their poor efficiency. This letter designs an efficient cross-lingual secure semantic searching scheme that performs semantic analysis, semantic matching, and relevance ranking on the ciphertext. The proposed scheme encrypts the language-aligned word vectors of the keywords, and further calculates the semantic similarity between the encrypted vectors to perform clustering for the encrypted keywords. Therefore, the scheme uses the encrypted language-aligned word vectors of query words to perform cross-lingual semantic matching only with the cluster centre keywords to get encrypted keywords semantically related to the query, instead of scanning all the keywords, thus significantly improving the search efficiency. The scheme also realises the ciphertext relevance ranking algorithm to rank the documents. Experimental results show the practicability of the proposed scheme.