Lightweight intrusion detection method and system of internet of vehicles based on knowledge distillation

Abstract

A lightweight intrusion detection method includes obtaining a feature data set of the internet of vehicles and pre-processing data; dividing pre-processed data into an initial training set, an initial verification set and an initial test set according to a preset proportion, performing data balance on the initial training set to obtain a balanced training set, performing feature selection on the balanced training set, the initial verification set and the test set; obtaining a teacher model by training with the model training set and the model verification set; using the teacher model, the model training set and the model verification set for distillation training to obtain a student model; testing a size and complexity of the student model and the performance of the student model, and saving the student model that passes the test as a lightweight intrusion detection model; and deploying the lightweight intrusion detection model.