Enseble learning-based technology malicious traffic detection

With the trend of network traffic encryption, traditional malicious traffic detection methods face significant challenges as they are unable to identify threats by analyzing the contents of encrypted data packets. Although modern detection methods based on traffic characteristics and deep learning techniques do not rely on packet content, the issue of class imbalance in traffic data further complicates detection efforts. To address these challenges, this paper proposes a malicious traffic detection technique utilizing ensemble learning and compares it with Random Forest and Deep Neural Network (DNN) models. Through experiments on the NSL-KDD dataset, the results show that the deep neural network model performs well in detecting malicious traffic, and the overall accuracy rate reaches 82 %. Notably, the DNN model achieves over 50% detection accuracy for attack types with smaller sample sizes. In addition, the EasyEnsemble ensemble learning model shows strong performance in addressing class imbalance issues. These results suggest that combining deep learning and ensemble learning methods effectively improves the accuracy of malicious traffic detection, providing an innovative detection strategy for the field of cybersecurity.