Multi-party privacy protection naive Bayesian model rapid training and lightweight prediction method

The invention discloses a multi-party privacy protection naive Bayesian model rapid training and lightweight prediction method, based on a homomorphic encryption scheme NewPai proposed in an ACSAC conference in 2021, in combination with an over-growth vector technology, a multi-vector data statistics scheme supporting high-dimensional data packaging is designed, and on the basis of the NewPai scheme, the multi-vector data statistics scheme supporting high-dimensional data packaging is designed. A distributed threshold decryption algorithm and an intermediate result protection method are designed, and a fast training protocol of a security enhanced naive Bayes model is realized. Compared with a prediction protocol designed by CARER, the non-interactive naive Bayesian model prediction protocol designed by the invention has the advantages that the calculation and communication overhead of a user side is reduced, the data of the user is prevented from leaving the local, the security of the user data is enhanced,