Multi-keyword Searchable Encryption With Privacy Protection In Medial Cloud

With the evolving of e-healthcare industry,the global medical industry becoming more digitized and virtualized.Many medical organization solved the needs of growing EHRs by using medical cloud which can store and manage electronic medical records in a unified manner.Encryption can protect EHRs dates in semi-honest cloud server,but the EHRs data no longer has the original characteristics after being encrypted.When medical institutions need some medical data,they cannot directly distinguish the required data in the ciphertext.It is an urgent problem to be solved that to ensure the privacy and security of EHRs data,and also can do fast and accurate multi-keyword search on large number of EHRs ciphertexts.Searchable encryption is the solution to this problem in recent years.Although a great deal of research has been done on searchable encryption^[1-5],there are still deficiencies in search efficiency,fuzzy search.To solve those problem,we propose two efficient searchable encryption schemes to increase the efficiency and flexibility.First,we propose a multi-keyword range searchable system which implements multi-keyword search based on secure Knn.It supports the connection keyword query with complex query structure and range search of the encrypted data quickly and accurately.The index and trapdoor are generated by using random partition matrix and random number to achieve the hiding of search mode.Finally,the scheme uses the matrix Hadamard product calculation to reduce the search index and search trapdoor match time.Finally,we compare the performance of the simulation experiment with the two schemes which using the traditional KNN scheme.The experimental results show that our scheme can significantly reduce the time of initialization,index and trapdoor generation part.Then,in order to realize an efficient fuzzy search,we propose a multi-keyword fuzzy searable scheme with privacy protection in cloud which is built by asymmetric scalar-product-preserving encryption?ASPE?and Hadamard product.For fuzzy search,we employ Word2vec which is the technique of machine learning to obtain correlation score between data and queries.To achieve more accurate document matching,we use the Tanimoto Coefficient to calculate the correlation scores of each word vectors and calculate the relevancy between each document by using TF-IDF.Finally,through simulation experiments,it is proved that the time of initialization,index and search trapdoor part are shorter when the number of segments is closer to ???.We evaluated the accuracy of several different ways of constructing word vectors.