Abstract: To solve the problem of too large storage overhead existed at fuzzy retrieval algorithm in cloud computing, a proxy encryption fuzzy retrieval algorithm based on local sensitive hashing is proposed. At first, this algorithm converts file keywords to bigram vector representation, and uses local sensitive hashing (LSH) functions to build initial index and query for all bigram vector form keywords. Secondly, the original index and query are encrypted separately to form the encrypted index and the trapdoor. Finally, it utilizes the inner product operation of index and query to realize multi-keywords fuzzy retrieval. Security analysis proves that the proposed algorithm is safe in the case of the known cipher text. Experimental results shows that improved algorithm can avoid the defects of large storage overhead due to excessive indexing by means of reducing the index, and effectively supports the multiple keywords retrieval. Its encryption and decryption performances are better than the contrast algorithm.