Harnessing Decision Tree-guided Dynamic Oversampling for Intrusion Detection

Imbalanced datasets present a significant challenge in the realm of intrusion detection, as the rare attacks are often overshadowed by the normal instances. To tackle this issue, it is essential to utilize the various strategies of imbalanced learning that aim to mitigate the effects of class imbalance and improve the performance of intrusion detection systems. One effective approach for dealing with class imbalance is through data augmentation methods like the Synthetic Minority Oversampling Technique (SMOTE). This research presents a novel data resampling approach that performs adaptive synthetic sampling on rare and complex samples by using decision boundaries. The benchmark dataset NSL-KDD was used to evaluate and validate the effectiveness of this approach. The experimental results demonstrated a significant improvement in the detection accuracy of rare classes, achieving 42% for u2r instances and 83% for r2l instances.