An Online Transfer Learning Model for Intrusion Detection using FT-Transformer and KSWIN-Driven Concept Drift Detection Mechanism

Deep learning has been successful in intrusion detection, but new attack methods cause data distribution changes, known as concept drift. To address this, we propose an online transfer learning method. It includes a concept drift detection mechanism to identify traffic pattern changes and fine-tune deep learning models accordingly. Experimental results show an average accuracy of 99.46% on the CIC-IDS dataset and 96.79% on the NetFlow V2 dataset.