Study On Social Discovery Based On Fast Search Of Encrypted Images

In recent years,more and more users like to use social media sites,and it is more popular that sharing photos on the sites,these photos can reflect their owner's interests.Generally speaking,the photos shared by users with same interests are very similar,so it is easy to expand users' circle of friends by recommend them for strange users or groups who may be have same hobbies with them according these photos,and it has brought new opportunities for the application of social discovery.However,the social network is also faced many difficult problems,such as the management of small amounts of data,the high cost of storage,the low reliability nature of the sites,and so on.Hence,the social networking sites tend to store amount of data on the cloud server to save the cost of the local storage and computing resources.While it is easy to leak users' privacy information for storing users' photos on the public cloud.Due to the privacy concerns,it is necessary carry on the encryption processing before outsourcing data to the remote server.When friends recommended function implemented,the operations on users' images should be conducted on ciphertext.Therefore,the research based on encrypted image of the social networking sites has important theoretical significance and application value.This paper mainly studies the social discovery based on the fast retrieval of encrypted images.The main work is as follows:Throughout this paper,we analyze a scheme of Yuan and others,and point out many shortcomings of the scenario,such as large storage space and low efficiency of that scenario.In view of these shortcomings,we put forward a social discovery scheme based on fast retrieval for encrypted images.Firstly,we provide the formalized definition and the privacy security requirements of this scheme,and then propose a privacy protected search scheme based on similarity computing.This scheme pre-build a searchable and encrypted index describing users' image,the index based on MD-algorithm,and it is stored on the cloud server with users' images,then through the computation of the secure similarity,we can find users with similar images,implementing friend recommended effectively.Compared with the existing schemes the proposed scheme can not only accurate to find a user whose image is most similar to that of other users,but also reduce the storage space from O(N)to O(logN),and the search time is also reduced from O(N)to O(logN),effectively reducing storage and time overhead.Finally,we do a simulation experiment,and the result of this simulation experiment confirms that the scenario we present can accurately find the users with similar images.