One of the most prevalent challenges nowadays is detecting intrusions into the Internet of Things (IoT) systems, which pose a variety of wide-ranging cyber threats. These devices encompass smart cities, industries, and homes, all integral to modern living. Their widespread adoption increases the urgency of addressing security vulnerabilities. Ensuring secure user interactions is of particular importance. This study proposes an intrusion detection approach that combines K-Nearest Niebuhr (KNN) and the Chimp Optimization Algorithm (ChOA) for detecting various and advanced cyber threats. Integrating ChOA with KNN aims to enhance classification accuracy by selecting the optimal subset of features from the dataset. To accomplish this objective, we devised the ChOA algorithm, employing the S-shaped and V-shaped models for conversion to the binary system. The algorithm incorporates three crossover operators and mutation techniques: average crossover, discrete crossover, flat crossover, and boundary mutation. These techniques significantly contribute to the population’s overall diversity and lead to the optimal solution through continued exploration and exploitation, resulting in high accuracy in identifying cyber threats. We evaluated the proposed approach on nine IoT-related datasets. We analyzed detailed performance metrics such as accuracy, precision, recall, and F1-score. IBChOA1 demonstrated superior results, outperforming the original ChOA and six other algorithms. We also compared our developed algorithm with six metaheuristic algorithms, demonstrating significant superiority in accuracy and fitness, and with four machine learning algorithms, demonstrating a notable edge over them. The enhanced ChOA outperforms standard methods due to its refined balance between exploration and exploitation, which is facilitated by S-shaped and V-shaped transfer functions, as well as robust feature space navigation enabled by diverse crossover methods and boundary mutation. These improvements ensure comprehensive feature selection, optimizing performance by effectively using the most predictive features. We evaluated it using several measures such as accuracy, recall, precision, F1 score, fitness function, and feature selection. We conducted a statistical analysis using the Fredman test and Wilcoxson test to statistically verify the results, highlighting the significance of the proposed approach compared to other algorithms based on accuracy. As a result of this study, we conclude that the proposed IBChOA1 greatly improves the security of IoT systems, providing a powerful way to find complex cyber threats. It is highly efficient in detecting diverse cyber-attacks, which has broader implications for the deployment of secure IoT networks in various fields, such as smart cities and industrial IoT.