Searchable Encryption Scheme And Application

With the development of various technologies in the Internet era,the size of data in the fields of scientific research,medical treatment,and education is exploding,but local devices can not meet the need of storage and computing.The powerful computing capability and storage space of the cloud attract everyone's attention,more and more users upload their own data to the cloud.However,there exist some security risks in the data stored in cloud devices.In order to solve the problem of cloud data security,searchable encryption merges at the right time.Attribute-based searchable encryption enables cloud servers to perform fine-grained searches of encrypted cloud data.However,in most schemes,access policies exist in the form of plaintext,which can easily reveal user privacy.By hiding the policy,the privacy can be protected to a certain extent,while the correctness of cloud search results cannot be guaranteed.And once the data user is authorized,the access to the data stored in the cloud will be unlimited.Most proposed searchable encryption schemes are applied to document search,and are not suitable for the retrieval of image content.Therefore,this thesis starts from the demand of retrieving image content,a searchable encryption scheme supporting image keyword retrieval is proposed.The research content of this thesis is as follows:1.Aiming at the problem that the existing searchable encryption scheme cannot ensure the correctness of cloud search results and limit user access,a verifiable multi-keyword searchable encryption scheme is proposed.The scheme uses the check key generated by the user as a condition for the cloud to determine the number of user access.The scheme introduces an authority to generate a trapdoor for the user and verify the correctness of the cloud search result,which reduces the computational cost of the user.Users send a partially hidden access structure to the cloud to implement policy hiding and multi-keyword search.The security and practicability of the scheme are proved by the definition of the security model and performance analysis.2.Aiming at the problem that the method of keyword extraction in searchable encryption scheme is widely used in documents,a novel image content recognition scheme based on convolutional neural network is proposed.The scheme introduces color features of the image to improve the sensitivity of the neural network to color.Combining the Bagging integration algorithm with the convolutional neural network reduces the complexity of the neural network model and effectively avoids overfitting of training data.The experimental results show that the recognition accuracy of the scheme is 98.44%,which is 3.59% higher than that of the convolutional neural network model,and has a good image recognition effect.3.Combining the above two research schemes,a searchable encryption scheme supporting image keyword retrieval is proposed.In this scheme,the proposed image content recognition model is used in the searchable encryption scheme to realize automatic extraction of image keywords,which improves the encryption efficiency of the data provider.Formal security proofs and simulation experiments show that the scheme is safe and practical.