Design And Implementation Of Classification Algorithm For Encrypted Data

In recent years,the rise of the subject of artificial intelligence has greatly promoted society towards the era of big data,It is increasingly required that we store data in the cloud for storage and calculation,Accompanying it is the leakage of sensitive private data.Nowadays,machine learning is being widely used in various fields of our lives.When training large-scale data from various channels,the effects of these machine learning models are more obvious and effective,massive data is often prone to privacy issues during the collection process.In particular,the medical field,the financial field,and the network security and other fields have extremely high requirements for data privacy.Ensuring data confidentiality and integrity has become an important challenge for our research.The existing machine learning algorithms we know cannot analyze and calculate the encrypted data,which has also promoted the research and development of machine learning in the field of encrypted data.Most machine learning algorithms cannot effectively process encrypted data,and the emergence of homomorphic encryption has brought new ideas.Through homomorphic encryption technology,we can directly operate on the ciphertext,and the result obtained is exactly the same as the result of the same operation we directly performed on the plaintext.In this paper,based on some existing work,combined with Paillier homomorphic encryption scheme and AES encryption scheme,three safe machine learning classifiers are proposed,including Bayesian classifier,decision tree classifier and hyperplane classifier.This paper studies several machine learning algorithms and combines the homomorphic encryption technology and AES encryption technology to construct secure machine learning classifiers.The main contributions and research results of this article are as follows:1Design and give a protocol for comparison on ciphertext encrypted by Paillier scheme,because Paillier encryption algorithm can only perform some basic operations on ciphertext,and does not support the operation of some machine learning algorithms.This protocol reduces the interaction between the two sides of the comparison times,which improves the efficiency of comparison.The comparison protocol is used to complete a secure hyperplane classifier construction scheme and a Bayesian classifier construction scheme,ensuring data privacy and ensuring the classification effect.2Design and give a decision tree classifier construction scheme in ciphertext.Under this scheme,we use the AES encryption scheme to encrypt the decision attributes to ensure the privacy of the data.We obtain a secure decision tree classifier model by training the ciphertext data.This scheme greatly guarantees the accuracy of the classifier.