Hybrid Deep Learning and Feature Engineering-Driven Adaptive IDS for Strengthening IoT Network Cybersecurity

Abstract

The rapid expansion of Internet of Things (IoT) deployments has greatly increased the potential attack surface, exposing these networks to a wide spectrum of cyber threats. Conventional intrusion detection systems (IDSs) often face difficulties in accurately identifying advanced attacks within the diverse and dynamic IoT environment. This study presents an adaptive IDS that integrates hybrid deep learning with feature engineering to improve IoT network security. The framework applies feature engineering to derive discriminative and relevant attributes from network traffic, which are then processed by a hybrid model combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) network. The CNN autonomously learns hierarchical representations from the processed data, while the LSTM captures sequential dependencies, enabling effective detection of intricate and evolving threat patterns. Evaluation on a publicly available IoT traffic dataset reveals that the proposed approach outperforms existing IDS solutions in terms of detection accuracy, precision, recall, and F1-score. Moreover, its adaptive capability allows for dynamic parameter tuning and feature selection in response to changing threat landscapes, ensuring consistent and reliable defense for IoT infrastructures.