Intrusion detection systems (IDS) are critical for safeguarding networked environments against evolving cyber threats. This paper presents an innovative approach for feature selection in IDS using the Superb Fairy-wren Optimization Algorithm (SFOA). The proposed method effectively balances exploration and exploitation through three unique behavioral phases, i.e., growth, breeding, and feeding, and predator avoidance behaviors of fairy-wren birds. Using UNSW-NB15, the performance of SFOA is evaluated against four established optimization algorithms: Cheetah Optimizer (CO), Harris Hawk Optimization (HHO), Whale Optimization Algorithm (WOA), and Parrot Optimization (PO), in terms of accuracy and selecting optimal feature subsets. Experimental results demonstrate that SFOA achieves the best classification accuracy and faster convergence while reducing the feature dimensionality. By balancing exploration and exploitation, SFOA enhances IDS efficiency, ensuring high detection rates and low false alarms for robust cybersecurity.