Research On Semantic-aware Multi-keyword Ciphertext Search Technology In Cloud Environment

Cloud computing is a business computing model that can provide Internet resources according to users’ needs.The application system deployed by the user on the cloud server can obtain the resources on the cloud according to the demand,such as computing power and storage space.This demand can be unlimited,and the user is only required to pay according to the amount.With the increasing development of cloud computing technology,more and more data owners choose to outsource their data to cloud servers.However,outsourced data is at risk of being leaked.In order to protect the privacy of outsourced data,the most common and direct method is to encrypt the data before outsourcing to the cloud server,and then outsource the encrypted data to the cloud server.However,the availability of encrypted data decreases,which will invalidate the traditional searchable technology.Traditional searchable encryption schemes are mostly based on the term frequency-inverse document frequency(TF-IDF)model.The TF-IDF model is based on frequency statistics and ignores the semantic information of the documents.How to design an efficient semantic-aware search scheme is a challenge.This thesis proposes two schemes for the problems of multi-keyword search over encrypted cloud data and the existing searchable encryption schemes:(1)Semantic-aware Multi-keyword Ranked Search Schemes over Encrypted Cloud Data(SRSE):In this thesis,we propose an accuracy-first semantic-aware multi-keyword ranked search scheme over encrypted cloud data(AF-SRSE)which adopts the LDA topic model and a clustering algorithm.We first present algorithms to generate a semantic-aware clustering document list by using the bisecting k-means algorithm.Then,we construct a binary tree index according to the document list.This scheme can effectively improve the search efficiency while ensuring the search accuracy.On this basis,we further transformed the index and proposed efficiency-first semantic-aware multi-keyword ranked search scheme over encrypted cloud data(EF-SRSE).We link leaf nodes with pointers,and adjacent leaf nodes have more relevant semantics,so that we only need to search the most relevant document and obtain a specified number of documents based on the pointers.(2)Semantic-Based Multi-keyword Ranked Search Schemes over Encrypted Cloud Data(SBERT-SRSE): Since the traditional searchable encryption schemes are based on the bag-of-words model to construct document vectors,the semantic information of the documents is ignored.In this thesis,we propose a privacy-preserving searchable encryption scheme based on the SBERT model.In this thesis,we use the SBERT model to train document vectors.Then,HDBSCAN is used to generate soft cluster member vectors for each document.Finally,the document topic vectors and keyword topic vectors are obtained.Through the SBERT model,our scheme measures the correlation between queried keywords and documents from the perspective of topic semantics,which can achieve more accurate semantic-aware multi-keyword ranked search over encrypted cloud data.(3)This thesis describes the specific requirements of the AF-SRSE scheme from the perspective of the practicability and achievability of the scheme.According to the design process of the prototype system of the AF-SRSE scheme and the different entities in each specific process(data owner,data user,cloud server)functions to implement the prototype system.Through relevant theoretical support and a large number of experimental results on real datasets,we confirm that scheme proposed can achieve the secure and efficient semantic-aware multi-keyword ranked search over encrypted cloud data.