KNN Classification Technology For Homomorphic Encryption

At present,the data security focuses on the whole life cycle security of data.Not only secure storage and transmission,but also focuses on the privacy protection in data usage in the cloud.The traditional encryption method can make sure that the safe storage of a large number of data in the cloud,but it can't guarantee the security of data in the process of using.Homomorphic encryption technology can be used in cloud computing to improve the security and confidentiality of data processing in the cloud because of the same characteristics of ciphertext calculation results after decryption and plaintext calculation results.Homomorphic encryption can ensure that the encrypted data can be processed without disclosing any information about data.Only the owner and the authorized visitors can decrypt the ciphertext by the private key,the participant cannot obtain the plaintext data.Other data privacy protection algorithms need more complex and tedious network transmission than homomorphic encryption technology,and face the problem of unauthorized access.A ciphertext comparison algorithm for the homomorphic encryption of the integer arithmetic operation(HE-IAO)has been proposed in this paper at first.The carry flag bit has been improved in binary subtraction,and then obtains the comparison result of ciphertext by homomorphic operation of carry flag and subtraction result without decryption.Compared with secure third-party based ciphertext comparison algorithm,HE-IAO based ciphertext comparison algorithm can effectively resist unauthorized data collection based on cloud,such as password cracking,memory leakage and so on.The experimental of KNN classification technology for homomorphic encryption results show that the computational efficiency of the ciphertext comparison algorithm proposed in this paper is about 50%higher than that of the ciphertext comparison algorithm based on secure multiparty computation.Secondly,based on the KNN(k-nearest neighbor,KNN)algorithm,this paper improves the KNN classifier to adapt to the homomorphic encryption.The original Euclidean distance is changed to square kernel function between the data set and the test set to avoid the square root operation in the process of ciphertext calculation.In the original scheme,the nearest K sample points are selected by sorting instead of the test data set as the center and R as the radius.It reduces the computation requirement in ciphertext state and improves the efficiency of the whole scheme.Finally,this paper proposed the scheme by the improved HE-IAO homomorphic encryption scheme and KNN algorithm,the simulation experiment is carried out to the CRKNN(Ciphertext Ring based K-Nearest Neighbor,CR-KNN).Experimental results show that the scheme can support classification while ensuring the classification accuracy by ciphertext.The classification accuracy only depends on the radius parameter r in ring based KNN classification.KNN classification algorithm based on HE-IAO scheme can reduce the computing pressure and storage pressure of users through cloud computing,ensure the data security in cloud computing,and solve the problem of privacy protection the data usage to a certain extent.