Hybrid Intrusion Detection System for IoT Against Adversarial Threats Using Intelligent RDLS Model

The Internet of Things (IoT) has revolutionized smart city solutions, enhancing operational efficiency and user convenience. Nonetheless, the integration of deep learning (DL) models for detecting cyber-attacks in IoT ecosystems introduces critical security concerns. Adversarial attacks exploit vulnerabilities in DL models, tricking them into making incorrect decisions. This paper investigates the application of adversarial training techniques to fortify DL models against such threats in smart city environments. The study generates new test data specifically designed to evaluate the model’s resilience to adversarial attacks. Additionally, the model undergoes retraining with a comprehensive dataset that includes both original and adversarial samples to bolster its defense mechanisms. By implementing adversarial training, the research aims to reduce the risks from sophisticated adversaries who attempt to manipulate data and compromise security within smart city frameworks. Experimental results indicate that the initial model achieved an accuracy of 99% on the DS2OS dataset. After retraining with the enriched dataset, which incorporated adversarial examples, the model’s accuracy slightly decreased to 98%. These results highlight progress in developing robust DL-based systems for detecting adversarial attacks, enhancing security for IoT-enabled smart cities. The findings contribute valuable insights into improving the reliability and safety of DL models used in complex smart city infrastructures.