Research on Intrusion Detection Method Based on Combination of Decision Tree and Dynamic Parameter Tuning Optimization

In the current cybersecurity environment, organizations and individuals continuously face threats from malicious attackers. To counter various network attacks, intrusion detection methods have become increasingly important. Currently, machine learning has achieved significant results in intrusion detection, but several challenges remain. First, due to the high dimensionality of network traffic data, the trained models often require excessive time to detect attack categories. Second, the extreme imbalance between normal and abnormal traffic data leads to inaccurate detection results for certain types of traffic. To address these issues, this paper proposes an intrusion detection method tailored for high-dimensional imbalanced network traffic data. The method employs logistic regression to handle dataset imbalance by expanding underrepresented categories, uses the Lasso regression model for feature selection to identify the most significant features, and finally combines decision tree with Dynamic Parameter Tuning Optimization for intrusion detection, thereby improving detection efficiency. The proposed method is validated on the CICIDS2017 dataset.