Sliced Bidirectional Gated Recurrent Unit with Sparrow Search Optimizer for Detecting the Attacks in IoT Environment

In an era characterized by pervasive interconnectivity facilitated by the widespread adoption of Internet of Things (IoT) devices across diverse domains, novel cybersecurity challenges have emerged, underscoring the imperative for robust intrusion detection systems. Conventional security frameworks, constrained by their closed-system architecture, struggle to adapt to the dynamic threat landscape marked by the continual emergence of unprecedented attacks. This paper presents a methodology aimed at mitigating the open set recognition (OSR) challenge within IoT-specific Network Intrusion Detection Schemes (NIDS). Leveraging image-based representations of data, our approach focuses on extracting geographical traffic patterns. We observe that the Recurrent Neural Network exhibits suboptimal classification accuracy and lacks parallelizability for attack analysis tasks. Our investigation concludes that the Sparrow Search Optimization Algorithm (SSOA) serves as a foundation for constructing an effective assault classification model. This research contributes significantly to the field of network security by emphasizing the importance and ramifications of meticulous hyperparameter tuning. It represents a critical stride toward developing IDSs capable of effectively navigating the evolving cyber threat landscape. In the experimental analysis of proposed model reached the accuracy and 0.963% respectively.