Machine Unlearning for the XGBoost Model with Network Intrusion Datasets
Researchers have developed XGBoost-Forget, a novel machine unlearning technique specifically designed for the XGBoost model when applied to network intrusion detection datasets. This approach addresses a gap in existing unlearning research, which primarily focuses on deep learning and image data. Evaluations on the IoT-23 and GeNIS datasets indicate that XGBoost-Forget can effectively remove data points while maintaining high predictive performance and offering significantly faster unlearning compared to full retraining. AI
IMPACT This research could enable more efficient and privacy-preserving updates for machine learning models used in cybersecurity.