Media Retrieval Supporting Privacy Protection In Mobile Cloud Computing Environment

With the development of mobile cloud computing and the popularity of smart devices,media retrieval,as one of the fastest growing technologies,has been widely used in various fields of society.The storage and computing requirements of large amounts of data have always been the driving factors for data outsourcing services.For example,enterprises use the cloud for media data storage to achieve effective sharing,retrieval and use of media data.Media retrieval brings convenience to people’s work and life,but it also inevitably brings some risks of privacy leakage.In the process of retrieving media data,the user needs to send the feature vector of media data to the cloud service provider.If the feature vector information is illegally used by the cloud service provider or stolen by a malicious attacker,the user’s query privacy will be leaked.How to efficiently retrieve media data under the premise of protecting the privacy of media data and query is a key problem to be solved in media retrieval in mobile cloud computing environment.However,the existing media retrieval privacy protection schemes have low retrieval accuracy and efficiency,and lack of security,which cannot meet the actual application requirements.In this paper,two privacy preserving schemes to solve the privacy preserving problem in two media retrieval methods,combined media retrieval and cross-modal media retrieval were proposed.(1)Aiming at the privacy-preserving problem of combined media retrieval in mobile cloud computing environment,this paper proposes a privacy-preserving combined media retrieval scheme,which solves the problems of retrieval accuracy,retrieval efficiency and security in the existing privacy-preserving image retrieval schemes.In order to improve the retrieval accuracy,this scheme uses the combination of image and text features as the query conditions,which can support the text semantics as the supplementary conditions of image retrieval,so it can better express the user’s search intention.In order to improve the retrieval efficiency,the k-means clustering algorithm and hierarchical navigation small world graph(HNSW)were used to construct the index of image feature vector.In order to achieve higher security,an enhanced Secure kNN algorithm supporting limited key exposure was designed to support the index encryption and retrieval of image feature vectors.Compared with the traditional Secure kNN algorithm implemented in the symmetric cryptosystem,ESLKE can prevent the untrusted user from disclosing the key,ensure the security of the key,and resist the same-closeness-pattern chosen-plaintext attack(IND-CLS-CPA).Security analysis and experimental results show that the proposed scheme can ensure higher security,and has higher search accuracy and efficiency.(2)Aiming at the privacy protection problem of cross-modal media retrieval in mobile cloud computing environment,this paper proposes a privacy-preserving cross-modal media retrieval scheme,which solves the problem of low retrieval accuracy and retrieval efficiency in the existing cross-modal retrieval schemes.In order to improve the retrieval accuracy,this scheme uses discrete latent factor model-based cross-modal hashing(DLFH)for training and prediction.In order to improve the retrieval efficiency,the unified navigable small world graph(UNSW)was used to build an index for the hash codes of images and texts.In order to protect the security of the owner’s image and text index,and the confidentiality of the user’s query vector,the proposed scheme uses additive homomorphic encryption and obfuscation circuit for index encryption and retrieval.Finally,the security analysis proves that the proposed scheme is secure,and the simulation results show that the proposed scheme has advantages in search accuracy and efficiency.