CROSS-DATASET VALIDATION OF A DEEP LEARNING–DRIVEN INTRUSION DETECTION FRAMEWORK FOR IOT SECURITY

The rapid proliferation of Internet of Things (IoT) devices has introduced significant vulnerabilities, making IoT ecosystems attractive targets for sophisticated cyber-attacks.Intrusion Detection Systems (IDSs) powered by deep learning have emerged as a promising defense mechanism; however, most existing models suffer from overfitting to specific datasets and lack rigorous cross-dataset validation.This study presents a deep learningdriven intrusion detection framework specifically designed for IoT security, validated across multiple benchmark datasets, including CICIDS2017, UNSW-NB15, and Bot-IoT.The framework integrates convolutional and recurrent neural architectures to capture both spatial and temporal traffic patterns, supported by advanced feature selection and normalization techniques.Experimental evaluations demonstrate that the proposed model achieves superior detection performance, with an average accuracy of 98.7%, precision of 97.9%, recall of 98.4%, and F1-score of 98.1% across diverse datasets.Notably, the framework exhibits strong generalization capabilities when trained on one dataset and tested on another, outperforming baseline machine learning and deep learning methods.The results confirm the reliability and robustness of the proposed IDS under heterogeneous IoT traffic conditions.This work contributes to the development of scalable and adaptive intrusion detection solutions, offering practical insights for securing real-world IoT networks against evolving cyber threats.