Multiple classification algorithm based on ensemble learning for intrusion detection

As one of the promising technologies for enhancing network security, network intrusion detection is crucial for accurately identifying various network attacks. To advance the intrusion detection capability, a multiple classification ensemble learning algorithm (MCELID) is proposed by analyzing intrusion data from diverse perspectives. Firstly, an individual learner combining graph convolutional networks and long short-term memory networks (GCN-LSTM) is constructed to extract both structural information and temporal correlations from intrusion data. Meanwhile, a weighted support vector machine (W-SVM) model is employed for handling multi-classification tasks, where the conditional probability is calculated to derive the probability vector of input samples. Finally, the intrusion detection result is determined by a soft voting mechanism that combines the classification results from the GCN-LSTM and W-SVM models. Extensive experiments conducted on the KDD-CUP 99 dataset demonstrate that the proposed MCELID algorithm outperforms other existing methods in terms of detection accuracy, particularly in the recognition of DOS and Probe attacks.