The enormous number of interconnected smart devices has transformed the traditional Internet into the Internet of Things (IoT). Currently, the IoT is indispensable across several sectors, including healthcare, industry, agriculture, transportation, smart homes, and smart cities. Intrusion detection systems (IDSs) are essential for safeguarding IoT systems from cyber-attacks. This work presents a robust wrapper feature selection approach based on the Crow Search Algorithm (CSA) to improve intrusion detection in IoT networks. The proposed CSA-based approaches utilize V-shaped (VBCSA) and S-shaped (SBCSA) transfer functions to transform continuous CSA into binary. Five IoT datasets containing authentic IoT traffic data are utilized to assess the proposed approaches. The experimental results show that the CSA method with the V-shaped transfer function works better than both the S-shaped-based CSA method and the genetic algorithm in a number of measures, such as accuracy, chosen features, optimal fitness, precision, recall, and F1-Score. VBCSA approach achieved an average accuracy of 0.99 across all five used datasets, demonstrating its efficacy in intrusion detection in IoT networks.