Research Of Content-based Multi-source Encrypted Image Retrieval Technology In Clouds

With the rapid development of network and multimedia technologies,image retrieval has been widely used in the fields of search engine,medical diagnosis,electronic commerce,copyright protection and so on.Due to the increasing popularity of cloud computing and cloud storage technology,people begin to outsource local images and image retrieval service to cloud.However,the cloud is insecure,leading to the risk of private information leakage in images that are outsourced to cloud.In order to protect the privacy of images,images can be encrypted before they are outsourcing.However,encrypted images makes image retrieval difficult.Existing researches put forward many different solutions to this problem,but all can only support the scene of single image data source.When there are multiple image data sources,the query user needs to communicate the decryption key with each image owner to obtain the unencrypted retrieval results,which will increases the communication overhead of the scheme.Therefore,the encrypted image retrieval under multiple image data sources remains to be further explored.This paper proposes a content-based multi-source encrypted image retrieval scheme in cloud.Each image owner will outsource their encrypted images and encrypted image features to cloud for storage and retrieval.The cloud will establish a retrieval index based on the encrypted image features provided by the image owner,and execute the encrypted image retrieval request for an authorized retrieval user,then return the encrypted retrieval results,thereby complete the encrypted image retrieval operation.Firstly,this paper uses the standard key stream and secure multi-party computing technology to encrypt images and image features respectively to ensure security.Image owners do not need to communicate image and image feature encryption keys,and the query use also need not to communicate the image feature encryption key and image decryption key with image owners which will decrease the communicate overload of the scheme and conform the secure application of multiple image owners.Secondly,the index for encrypted image retrieval is established to improve the retrieval efficiency,and a new image similarity measure method is proposed to avoid the cloud obtaining the real similarity information between images during the retrieval process.Finally,theoretical analysis proves the security of the proposed scheme,and the experimental evaluation results based on the real image datasets demonstrate the effectiveness of the proposed scheme.