In dynamic traffic networks, intrusion detection systems (IDS) must handle dynamic data stream where traffic changes occur, and concept drift is customary. Traditional concept drift detection approaches often experience a challenge between sensitivity and stability, resulting in delayed adaptation and uncontrolled false alarms. This paper proposes an AdaptiveDelta ADWIN framework that tunes sensitivity detectors using online lightweight controllers: Volatility (VC), that tune a delta based on error volatility, and AlertRate Controller (ARC), which modulates the drift alarms frequency. The framework is implemented using Bagging ensemble of Hoeffding Adaptive Trees and evaluated on a network preprocessed traffic dataset. Comparative experiments opposed to a fixed, ultrasensitive delta detector illustrate that adaptive tuning authorizes timely drift detection while maintaining stability, decreasing false alarms by more than 25%, and enhancing predictive overall performance. AdaptiveDelta baseline maintains a stable accuracy approximately 80% 82% accentuating the importance of balancing detection sensitivity with operational stability in streaming IDS implementation. These results highlight the practical value of the proposed framework, which is lightweight, computationally efficient, and suitable for real-time deployment in streaming IDS environments.
Rodney Buang Sebopelo (Tue,) studied this question.