Semantic-aware Multi-keyword Ranked Search Scheme Over Encrypted Cloud Data

With the development of cloud computing technology,an increasing number of users choose to outsource their data to the cloud to save the cost of hardware and software maintenance.In order to protect the privacy of the outsourced data,data owners must encrypt their data before outsourcing to the cloud.However,the encryption can affect the usability of the data and the existing search schemes on plaintext data cannot directly applied to the ciphertext.It is a necessity to propose an effective searchable encryption scheme on documents.Recently,many searchable encryption schemes adopt the term frequency – inverse document frequency(TF-IDF)model.But TF-IDF model ignores the latent semantic features of the user's queried keywords and documents,which could lead to unsatisfactory search results.To solve the ignorance of semantics in searchable encryption schemes,this paper proposes two different schemes by adopting different semantic models and schemes.(1)Semantic-aware multi-keyword ranked search scheme over encrypted cloud data based on Doc2 Vec model(DMRSE): Based on the Doc2 Vec model,the scheme uses the Doc2 Vec model to extract the semantic features from the documents and generated the feature document vector.This feature vector is a low-dimension distributed representation.The scheme improves the search efficiency and space usage by adopting the low-dimension feature vector.Moreover,this scheme can support dynamic update on the document set,which largely improves the usability of this scheme.(2)Semantic-aware multi-keyword ranked search scheme over encrypted cloud data based on LDA topic model(LDA-MRSE): Based on the LDA topic model,this scheme extracts the latent topic information from the document set and generate the document-topic matrix and keyword-topic matrix.LDA topic model is an unsupervised learning model,which is more suitable to implement in cloud.To further improve the efficiency of this scheme,a special complete binary tree based index is proposed,which can achieve sub-linear search time cost.(3)Enhanced semantic-aware multi-keyword ranked search scheme based on LDA topic model and keyword extraction(LDA-ESSS): This scheme extracts the feature keywords from the dataset by the topic information,and extends the original topic vector with the feature keywords information.And this scheme further improves the semantic search accuracy of LDA-MRSE.This scheme solves the ignorance of queried keywords in LDA-MRSE.The experiment result shows that this scheme can improve the search accuracy and rank privacy without a significant decrease in search efficiency.