The Internet of Things (IoT) has brought billions of devices into a single network and has changed modern living. However, there are significant security challenges in such resource-constrained systems for which efficient solutions are necessary. This paper investigates applying advanced feature selection techniques such as K-Best, Recursive Feature Elimination (RFE), and BORUTA to enhance Intrusion Detection Systems (IDS) for IoT environments. We employed the CIC IoT 2023 dataset to compare the effectiveness of these methods in reducing the computational complexity of the model without jeopardizing the accuracy and reliability of the model. It reveals that reducing the number of features to a specific set enhances the model's efficiency without affecting the detection capability. These findings indicate the necessity of lowering features in developing very efficient but low-complexity IDSs that are important for developing IoT security. Future work will involve studying hybrid feature selection strategies and replicating these studies in different IoT contexts.