An Adaptive Hybrid MFO–TLBO Framework for Multi-Objective Optimization in Network Intrusion Detection Systems

This preprint presents an adaptive hybrid MFO–TLBO optimization framework integrated with Extreme Learning Machine (ELM) for network intrusion detection. The proposed method combines the exploration capability of Moth-Flame Optimization (MFO) with the exploitation strength of Teaching-Learning-Based Optimization (TLBO) to enhance feature selection and classification performance. Experimental evaluation on the KDDCUP1999 dataset demonstrates that the proposed hybrid approach achieves 95.19% detection accuracy, outperforming several standalone optimization-based IDS models. This version represents a preliminary preprint. An extended version including evaluation on modern datasets such as CICIDS2018 will be submitted for journal publication.