Development of Hybrid Intrusion Detection Systems for IoT Enabled Devices Utilizing Resource Constraint Learning Frameworks

The increasing prevalence of IoT-enabled devices in resource-constrained environments has made them vulnerable to various network intrusions and security threats. Traditional intrusion detection systems often struggle with capturing complex temporal dependencies in network traffic and balancing computational efficiency with accuracy, limiting their effectiveness in IoT networks. To address this challenge, we recommend a novel hybrid intrusion detection framework based on a hyperparameter-tuned LSTM-GRU model. The proposed system leverages the strengths of both architectures to efficiently analyse sequential patterns and temporal dependencies in network traffic while maintaining computational feasibility for IoT devices. A grid search method is employed to elevate hyperparameters like learning rate, batch size, and the number of layers, ensuring the model’s performance is fine-tuned for the intrusion detection task. Experimentation outcomes on the CICIDS 2017 dataset depict the model’s robustness, achieving 95% accuracy, 96% precision, 95% recall, 96% F1-score and 95% specificity, significantly outperforming existing approaches. This hybrid architecture offers a practical solution for real-time intrusion detection in IoT systems, balancing computational efficiency with high detection accuracy.