Semi-supervised Campus Network Intrusion Detection Based on Knowledge Distillation

With the rise of the concept of smart campus, cloud computing, Internet of things, augmented reality, and artificial intelligence are widely deployed on the campus. The campus network supports the above technologies and plays an important role in constructing a smart campus. At the same time, the campus network is also facing various security challenges. This paper proposes an intrusion detection method to ensure the security of the campus network according to the characteristics of the campus network. Based on the idea of edge computing, this paper deploys the intrusion detection model in a distributed manner at the network aggregation layer. To solve the problem that network intrusion labeled training data is difficult to obtain, this paper uses a semi-supervised method to train the detection model and uses knowledge distillation for model compression to reduce the complexity of the edge detection model. Furthermore, this paper proposes a model update strategy using samples from edge nodes. The experimental results show that the proposed method can obtain good detection results.