Network Intrusion Detection Method Based on Semi-Supervised Learning and Random Forest

Network Intrusion Detection Systems (NIDS), as critical tools for defending against malicious attacks, still have several limitations, such as false positives and false negatives, limited adaptability to new attacks, and scalability issues. To enhance the adaptability and real-time detection capabilities of NIDS for novel attacks, this paper proposes a network intrusion detection method based on semi-supervised learning and random forest. Firstly, an autoencoder is used to reduce the dimensionality of the NSL-KDD dataset, increasing robustness against high-dimensional complex data. Sencondly, a semi-supervised learning framework is established through the DBSCAN clustering method, improving the model's performance and stability. Finally, multi-layer perceptron (MLP) is employed to select data features and extract an optimal feature subset, and random forest (RF) classifier is constructed to reduce computational complexity and effectively boost model performance. Experimental results show that the proposed method improves accuracy on the NSL-KDD and UNSW-NB15 datasets by 8.92% and 6.1%, respectively, which demonstrates strong generalization ability and robustness, achieving precise and efficient detection and classification of network intrusions and attack types.