Traditionally, Wireless Sensor Networks (WSNs) lack built-in algorithms to identify and react to intrusions and threats, necessitating the establishment of Intrusion Detection Systems (IDSs). In practice, it is difficult to design an IDS in WSNs due to the large scale, mobility, and limited memory of the sensors in WSNs. Feature Selection (FS) responds to these challenges by reducing data dimensionality and improving IDS accuracy, thereby enhancing intrusion detection classification by selecting the most expressive features in intrusion detection datasets. This paper presents a new intrusion detection framework called the Enhanced Binary Aquila Optimizer (EBAO) model. EBAO aims to efficiently improve the feature space to enhance detection accuracy while minimizing computational complexity in IDSs. EBAO integrates four improvements in the original Aquila Optimizer (AO) to address the FS problem. First, it uses a hybrid initialization approach that combines the Lévy flight generation function and the random uniform generation function to generate suitable solutions for the FS problem. Second, it employs the β-hill climbing algorithm as a local search method to enable the AO method to search efficiently in the FS solution space. Third, it employs the mutation equations of the Harris Hawks optimization method in the optimization process of AO based on a probabilistic function to explore the FS solution space. Lastly, it models the FS solution space in AO using two categories of binarization techniques (S-shaped and V-shaped). The performance of EBOA was evaluated using ten WSN datasets and eight transfer functions and then compared with eight metaheuristic-based IDSs and six machine learning algorithms. The Wilcoxon pair signed rank and Friedman tests were used to determine the statistical differences and rankings of the evaluated algorithms based on classification accuracy and fitness value. The experimental and statistical analysis strongly indicated that EBAO demonstrates superior effectiveness compared to the eight popular optimization wrapper algorithms for all ten datasets, which highlights its robustness and reliability. EBAO is https://github.com/drnooraldeen/EBOA.git