Ates, G.Celebi, B.Semerci, U.A.Capkan, E.Yildirim, B.Ar, I.Arsan, T.2025-12-152025-12-1520259798331537197https://doi.org/10.1109/BlackSeaCom65655.2025.11193891https://hdl.handle.net/20.500.12469/7663"OB-IDS (Optimized BERT-based Intrusion Detection System)"presents a lightweight IDS optimized for the resource-constrained environments. The proposed system which employs quantization, pruning, knowledge distillation and self distillation methods is developed in a multi-stage optimization manner. It was evaluated on UNSW-NB15 and CIC-IDS2017 datasets. The proposed system was found to significantly reduce the inference time and memory usage while maintaining high classification accuracy. The experimental results demonstrate the applicability of Transformer-based IDSs for real-time threat detection in resource-constrained environments. © 2025 IEEE.eninfo:eu-repo/semantics/closedAccessBertCybersecurityFine-TuningIntrusion DetectionMachine LearningModel OptimizationNetwork SecurityQuantizationConference Object10.1109/BlackSeaCom65655.2025.111938912-s2.0-105021003093