The increasing scale and heterogeneity of internet of things (IoT) networks have intensified the need for accurate and computationally efficient intrusion detection mechanisms capable of handling high-dimensional and dynamic traffic data. Conventional intrusion detection systems (IDS), including deep learning-based approaches, often face challenges related to feature redundancy, optimization complexity, and adaptability in resource-constrained IoT environments. To address these limitations, this paper proposes a novel hybrid optimization-driven intrusion detection framework, termed LOA–QBSA–IDS, which integrates the Lion Optimization Algorithm (LOA) and the Quaternion-based Backtracking Search Optimization Algorithm (QBSA) for effective anomaly detection in IoT networks. In the proposed framework, LOA is employed as a feature selection mechanism to identify an optimal subset of discriminative features from IoT traffic data by modelling the social hierarchy and cooperative hunting strategies of lions. This process innovation significantly reduces feature dimensionality while preserving critical intrusion-related information. Subsequently, QBSA is utilized to optimize the classification process using quaternion-based representations, which enhance the search space exploration and convergence stability, thereby improving anomaly classification accuracy. The synergistic integration of LOA and QBSA enables efficient handling of complex optimization problems inherent in IoT intrusion detection. The proposed LOA–QBSA–IDS outperforms a state-of-the-art Deep Learning-based IDS (DL-IDS), achieving a 0.6% improvement in detection accuracy, along with improved robustness in anomaly identification. The results validate the effectiveness of the proposed hybrid optimization approach and highlight its suitability for real-time and resource-constrained industrial IoT security applications.
Santhanakrishnan et al. (Tue,) studied this question.