Research On Efficient Search Technology Over Encrypted Images In Cloud Computing

With the wide application of electronic acquisition technology on network communication devices,the visual resources such as images are increasing explosively.Traditional image search technology can not meet the needs of the times.As a sharable virtual resource pool,cloud computing can provide powerful computing and unlimited storage resources.To reduce software management and hardware maintenance,more and more people tend to outsource the huge amount of images to the cloud.However,due to untrusted cloud servers,the user needs to encrypt their images before uploading to prevent privacy leaks,which makes it more difficult to search images efficiently.Therefor,how to effectively retrieve large scale encrypted images in cloud computing is an urgent problem to be solved.Now,towards efficient privacy-preserving image search becomes a new research interest in academic community.Specifically,there are existing encrypted image search schemes having two defects in search performance.On one hand,the existing search system over ciphertext cannot get coincident accuracy to plaintext.Although,homomorphic encryption-based image search schemes could achieve comparable accuracy.They have the shortcomings of expensive computing and heavy communication.On the other hand,linear search should sort all data based on the similarity,which consume excessively long times.To solve the reality problems of high dimension in feature vector and large database,high effective index technique has important research significance.Therefore,we proposed an efficient encrypted image search scheme based on secure modular hashing.This paper has been accepted by International Academic Journal: Soft Computing(SCI Impact Factor: 2.472).The main contributions of this paper are two-folds:1.We propose a new efficient privacy-preserving image search over encrypted database in cloud computing.By incorporating K-means and secure modular hashing,the proposed scheme can protect the privacy of image features and achieve analogous distance-calculation under the ciphertext.The proposed encrypted image search scheme adopts a light-weight encryption to encrypt feature vectors,which makes the complexity over ciphertext be consistent with the search scheme over plaintext.Besides,comparing to the linear search,tree index can significantly improve search efficiency with high search accuracy.Furthermore,we provide a ASPE-based encrypted image search scheme without key sharing through taking advantage of asymmetric scalar-product-preserving encryption,which gives more protection to the secret key.2.We conduct a comprehensive security and efficiency analysis on the proposed solution.We also provide a detailed efficiency and accuracy evaluation over two realworld data sets: INRIA and Corel.Through the whole experiment,our scheme have obvious advantage in encryption time.Beyond that,by adjusting the length of ciphertext,it brings search efficiency and accuracy into balance.Last,based on two different similarity measures,the experimental results indicate that our system can achieve a comparable search accuracy.