Research On Ciphertext Decision Tree Classification Scheme Based On Functional Encryption

In the era of rapid development of artificial intelligence technology,machine learning brings various intelligent services to human.Human use machine learning to process data,but data information can't be effectively protected.How to use encryption algorithm to ensure the confidentiality of data is a problem worthy of study.As a machine learning algorithm,decision tree algorithm has the characteristics of simple,good fault tolerance and high robustness.The traditional encryption algorithm can't calculate the ciphertext and can't be applied to the decision tree algorithm.The computation of ciphertext is realized by fully homomorphic encryption or functional encryption.Therefore,in this paper,the ciphertext decision tree based on functional encryption is studied,the specific work is as follows:1.In the encryption algorithm,for the common homomorphic encryption algorithm has the problems of low efficiency and poor security.The FV homomorphic encryption algorithm based on RLWE is used.In this algorithm,there is no need of analog-to-digital conversion technology to manage noise,and a smaller calculation key is used.Its security depends on RLWE assumption.Based on FV homomorphic encryption algorithm,a functional encryption algorithm FVBFE is constructed.The advantages of the functional encryption are to calculate ciphertext data and return the calculation result of plaintext,which makes the algorithm get a good application in ciphertext decision tree classification algorithm.2.FVDT scheme is constructed by FV homomorphic encryption and decision tree algorithm.In order to solve the problem that the data of ciphertext decision tree algorithm may be real numbers,the method of analog operation,approximate value and scale factor is used to convert real numbers into integers.For the homomorphism calculation of ciphertext,the algorithm of ciphertext data comparison and the algorithm of maximum value of ciphertext data are designed.The scheme is used to complete the classification of ciphertext decision tree,and SIMD technology is combined with the scheme to improve the efficiency of ciphertext decision tree classification.3.FVBFE_DT scheme is constructed by functional encryption scheme and decision tree algorithm.For three different decision tree classification algorithms ID3,C4.5 and CART,three ciphertext decision tree classification algorithms are constructed by functional encryption.The advantage of this scheme is that functional encryption can compare ciphertext data and get plaintext comparison result directly.In the process of ciphertext data comparison,users do not need to interact with ECS,the users do not need to keep online all the time,reducing the cost of communication.The FVDT scheme and FVBFE_DT scheme constructed in this paper can meet the security requirements in practical application.The experimental data show that the two schemes constructed in this paper are better than Bost's scheme in efficiency.