Application And Research On Privacy-preserving Image Retrieval In Cloud Environment

With the rapid development of 5G communication and Internet technology,content-based image retrieval is widely used in various fields,but also faces the challenge of mass data storage and analysis.In order to reduce the storage and computing burden on the client side,images are often outsourced to a cloud server.However,because the cloud server is not completely trustworthy,if the image is directly uploaded to the cloud,it may violate the user’s facial features,geographical location and other personal privacy information,especially in the field of investigation.Privacy preserving image retrieval can alleviate the above problems.By encrypting image features and outsourcing them to the cloud,it not only protects the content privacy of images,but also reduces the local computing cost.Therefore,for clients with limited resources,this paper studies the image retrieval scheme for privacy preserving in the cloud environment.The main work is as follows:(1)Aiming at the problem of low efficiency and accuracy of encrypted image retrieval,a privacy-preserving image retrieval scheme based on deep convolutional neural network(DCNN)and vector homomorphic encryption(VHE)is proposed.In the field of electronic investigation,the ciphertext-policy attribute-based encryption is used to set the access policy,and the DCNN and Hash are used to obtain higher quality features.The above methods improve the security of users and the accuracy of retrieval.The features are encrypted in batch by vector homomorphic encryption and outsourced to the cloud to alleviate the local encryption overhead.A new outsourced K-means clustering algorithm is designed to construct an encrypted index tree in the cloud,which greatly reduces the retrieval time and improves the retrieval accuracy.Security analysis shows that the scheme not only achieves the confidentiality of encrypted data,query and returned results,but also hides the user’s identity.(2)Aiming at the problem of high local computation overhead,a lightweight image outsourcing retrieval scheme based on attribute-based encryption is proposed.The scheme is a multi-user oriented lightweight ciphertext image retrieval scheme,which realizes secure and efficient ciphertext image search.The proposed scheme improves the encryption model based on color features and extracts ciphertext features through the cloud server,which reduces the computational overhead of the user.For multi-users,the trusted third party and proxy server are introduced to perform secondary encryption and identity verification and the multi-user access policy is set,which can resist the collusion attack of users and enhance the security of the system.On this basis,users outsource the pre-decryption task to the proxy server,which reduces the local computing overhead and provides effective privacy protection.(3)An electronic investigation image retrieval system is designed and implemented through Matlab platform,which increases the application scenarios and further verifies the effectiveness of the scheme proposed in this paper.The specific functions of the system include user registration and login,access control,image encryption,index establishment,image retrieval,image decryption.The effectiveness and security of the system are proved by the experimental simulation test of ORL face database,which further proves that the electronic investigation image retrieval system has practical application value.