Abstract
Severe class imbalance in large-scale Information Systems data environments, such as those generated by IoT infrastructures, can significantly degrade the performance of intrusion detection systems (IDS) in identifying rare yet critical security threats. While existing oversampling techniques generate synthetic minority-class data to mitigate imbalance, they often introduce substantial computational overhead that limits system efficiency and scalability, a critical gap between model effectiveness and deployability in real-world IS infrastructures (Kovács, 2019). We propose a computationally efficient class resampling strategy that selectively targets underperforming minority classes rather than uniformly resampling all classes to enable scalable, cost-efficient intrusion detection, supporting real-time threat detection in large-scale IS environments.
Recommended Citation
Nadim, Mohammad; Chennamaneni, Anitha; and Gupta, Babita, "A Selective Class Resampling Strategy for Robust Intrusion Detection Systems Under Severe Class Imbalance" (2026). AMCIS 2026 TREOs. 77.
https://aisel.aisnet.org/treos_amcis2026/77