Research On Key Technologies Of Homomorphic Encryption

The continuous advancement of science and technology has driven the rapid development of the information industry.Currently,the processing method of cloud data for user's data is generally that the user uses the cloud computing platform to transmit the data storage to the cloud,and the cloud service providers encrypt the data of the user.When the user uses the data,the encrypted data is retrieved and decrypted,and the cloud service provider cannot change the ciphertext,otherwise it cannot be decrypted into the correct plaintext.At the same time,cloud service providers are not completely trusted.In recent years,private data leakage incidents have occurred frequently.The best idea is to make full use of the powerful computing power of cloud service providers to process ciphertext,just like the corresponding processing of plaintext.That is to say,ciphertext can be directly processed on the cloud and can be decrypted into the corresponding processed plaintext.This is the concept of homomorphic encryption.Although the homomorphic encryption algorithm has made significant progress in construction,the computational overhead of homomorphic encryption is too large,which restricts its practical process.Although there are many optimization and improvement algorithms to promote the research process of homomorphic encryption,there are still some shortcomings such as excessive public key overhead and low algorithm efficiency.To this end,this doctoral dissertation takes the problem of excessive public key overhead in the process of homomorphic encryption calculation as the main research goal.On the other hand,the current situation of big data security and privacy in practical applications in the cloud computing environment is not optimistic.At the same time,this doctoral dissertation also conducts indepth research on the problems that need to be solved urgently in the application of network coding pollution attack,digital image privacy disclosure and privacy protection of machine learning.The main research results are as follows:1.Aiming at the problem that the public key overhead of the homomorphic encryption algorithm on the integer is too large,a short public key full homomorphic encryption algorithm based on PAGCD is proposed.Firstly,a partial homomorphic encryption algorithm is constructed,and the algorithm is self-lifting by using the compression and decryption circuit technology to construct a full homomorphicencryption algorithm on the integer.Convert the public key into a quadratic form,reducing public key overhead without semantic security.Theoretical analysis shows that the proposed algorithm has smaller public key size than the original original algorithm,and the efficiency of the algorithm is significantly improved.2.Aiming at the problem that data is vulnerable to pollution attack in network coding,a network coding anti-pollution attack scheme with mixed homomorphic signature is proposed.Firstly,the source node set,non-source node set and link set of directed multi-graph are used to model the wireless network coding process,and the network anti-pollution model is established considering the two types of pollution attacks: data pollution attack and label pollution attack.Secondly,a hybrid homomorphic signature scheme is established by using message authentication code,dynamic message authentication code and homomorphic signature scheme to improve the message verification process of anti-pollution attack model,ensure the integrity of each message encoding packet content and improve the security performance of the algorithm.Finally,the safety performance of the scheme is verified by comparing the percentage of polluted nodes,the cumulative distribution of traffic and the calculation efficiency under the experimental simulation environment based on the coding and transmission mechanism of adaptive security network.This doctoral dissertation proposes a new way of thinking about the problem that the coded packet is easy to be stolen,tampered and polluted in the anti-pollution attack of wireless network.3.Aiming at the problem that medical images are easily leaked in network transmission,a medical image information hiding scheme based on full homomorphic encryption is proposed.The basic principle is that the ciphertext image obtained by the XOR Scramble encryption method is first used,and then the original medical image pixel classification bit stream and the user's private information are homomorphically encrypted,and their ciphertexts are homomorphically operated and embedded in the encrypted image to obtain a secret image.In order to ensure that the receiver obtains a complete hidden image,the network anti-pollution attack model designed in this doctoral dissertation is used for transmission.The experiment results show that the proposed scheme can satisfy the privacy protection and the integrity of medical images.4.Aiming at the problem of medical image privacy leakage in cloud computing environment,a privacy-protected full homomorphic encryption scheme based on extreme learning machine is proposed.Under the condition of general ciphertextretrieval model,a four-part participant model consisting of user,encryption server,ciphertext index server and analysis server is designed.Considering the data integrity,the user uses the network anti-pollution model designed in this doctoral dissertation to transfer data to the encryption server.The experimental results of training medical image datasets show that compared with other schemes,the privacy protection limit learning machine proposed by this scheme has higher accuracy.This doctoral dissertation proposes a corresponding solution to the key problems such as network pollution attack,medical image privacy leakage and machine learning privacy protection by improving the full homomorphic encryption algorithm on integers and using improved algorithms.Theoretical analysis and experimental results show that under the network anti-pollution attack model studied in this doctoral dissertation,the existing evaluation standard and the improved homomorphic encryption scheme are used in medical image privacy protection and medical image ciphertext retrieval applications with good results.