Privacy-preserving k-nn classification method based on vector homomorphic encryption

Abstract

The present invention relates to a privacy protection K-NN classification method based on vector homomorphic encryption, comprising: A. receiving a query vector group and a standard vector group; B. generating a matrix G through the query vector group, and the standard vector group uses the key S through State encryption to generate a ciphertext group and a new key GS; C. Convert the new key GS to a conversion key S' to obtain the conversion matrix M and conversion ciphertext group at this time; D. Use the conversion key S 'Decrypt the converted ciphertext group to obtain the decrypted vector group; E. Attach classification labels to the corresponding query vectors according to the components of the K minimum decrypted vectors. The invention can protect the user's private data well, and when the private data is protected, the user's query vector can be efficiently and accurately classified through the K-NN algorithm, which improves the efficiency of vector type judgment and expands the vector The scope of application of type judgments.

Description

technical field [0001] The invention relates to a vector classification method of a K-NN algorithm under privacy protection, in particular to a privacy protection K-NN classification method based on vector homomorphic encryption. Background technique [0002] K-Nearest Algorithm (K-NN) is widely used in pattern recognition, and also has good performance in data classification. In the era of big data, complex calculations are often outsourced to third-party clouds. However, in this process, the outsourced data may contain some sensitive data of sensitive users. Directly outsourcing user information to a third-party cloud may cause some private information of users to leak out, causing losses to users. A direct solution is to encrypt the data and then upload it to the cloud, but this also brings some new challenges to the efficiency of the K-NN algorithm. Homomorphic encryption technology can operate under ciphertext, and the result after decryption is correct. Therefore, t...

AI Technical Summary

Problems solved by technology

Due to the K-nearest neighbor rule for probability estimation density, there are two disadvantages. One is that the integral of the estimated density must be diffused to infinity, and the other is that there is a heavy burden on the calculation.

Images