Research And Application Of Homomorphic Encryption Algorithm In CryptDB

With the advent of the Big Data era,the proliferation of data volumes has led to the widespread use of databases in a variety of new scenarios,especially when data is stored in third-party storage areas that are not under the control of the user.Privacy data is exposed to serious external attacks and internal leaks Threat,and currently there is no mature ciphertext database product to solve these problems.MIT developed the CryptDB system in 2011,and although the product is still experimental,its innovative way to build encryption has been borrowed from Google and other companies.This paper focuses on the CryptDB system used in the homomorphic encryption algorithm,respectively Paillier algorithm and batch order-preserving encryption algorithm.Paillier algorithm is a probabilistic public-key encryption algorithm.In this paper,we propose to use the Chinese residual theorem to reduce the scale of the modular operation and to precompute the partial modular exponentiation to be used for actual encryption.The mode size in the operation is reduced by half and the part of the ciphertext message The number of modular multiplications decreases from log2m to 1.In terms of security,leakage of intermediate operation results from China's residual theorem leads to the key being unsafe.However,both the calculation process and the result are generated and stored in the trusted CryptDB agent together with the encryption key.The calculation using a public key does not affect the overall security.Experimental results show that the improved Paillier algorithm can increase the average encryption speed by 7.81ms.After applied to CryptDB,the response time to insert and update of different data amounts is shortened by 7.42%and 12.14%on average,and the average throughput increases 8.04%and 19.70%,With different types of indexes,the average response time is shortened by 7.74%and the throughput is increased by 8.2%on average.The order encryption algorithm is used by CryptDB to re-encrypt data after the user goes online.In this paper,we propose to replace the AVL tree with a scapegoat tree to construct the plaintext space used in the algorithm encryption,and do not need to rotate the subtree frequently in the process of establishing the search tree.Instead,we use the time complexity of O(logM)to solve the unbalanced subtree reconstruction.In terms of security,although a small amount changes the plaintext space arrangement and storage method,the algorithm does not store the plaintext space after the encryption and decryption are completed,and does not affect the IND-OCPA.Experiments show that when the parameters of scapegoat tree are taken as 0.75,the encryption rate of AVL tree is increased by 31.89%compared with that of BLA algorithm.After applying to CryptDB,online users with different amounts of data reduced the average online time by 13.1%.CryptDB system in the application of improved encryption algorithm,and build cloud platform environment for functional testing and performance testing.First of all,in functional testing,database operations and platform business tests in two ways,that CryptDB support basic database operations,data encryption and decryption work,and data authentication and access control,even unauthorized administrators can not view user data.Then,in the performance test,we test the performance of the CryptDB and Mysql programs before and after the improvement by using business processes such as platform virtual resource information storage,resource agent online and publishing tasks.Experiments show that with different amounts of data and different concurrent database requests,the amount of database requests for CryptDB support increased by 28%,the amount of data supported by users increased by 23.04%,the number of task requests supported by the platform increased by 29.31%.Finally,the development of database testing tools from different levels of transaction operation of the database performance experiments.Experiments show that the improved CryptDB transaction response time is still several times the difference with the MySQL database,but an improvement of 10.25%before the average,to meet the real cloud platform backend requirements.In this paper,two encryption algorithms with homomorphism used in CryptDB system of ciphertext database are optimized in terms of algorithm implementation.Experiments show that the improved scheme can optimize the performance of CryptDB system and will continue to explore the performance of CryptDB Bottleneck,with a view to further improve the database user experience.