Intrusion Detection in IoT Using Deep Recurrent Neural Networks: A Complex Network Approach to Modeling Emergent Cyberattack Behaviors

The rapid proliferation of Internet of Things (IoT) infrastructures has introduced significant security challenges due to device heterogeneity, dynamic interactions, and resource limitations. Traditional intrusion detection systems (IDSs) often struggle to capture temporal dependencies and emergent behaviors inherent in modern IoT cyber threats. This study presents a novel hybrid framework that combines deep recurrent neural networks (RNNs), specifically long short-term memory (LSTM) architectures, with complex network modeling to enhance the detection and classification of sophisticated attacks. The proposed system leverages normalized and labeled IoT traffic data, encompassing multiple attack classes (e.g., DoS, DDoS, Brute Force, MITM, and Replay) to train an LSTM-based IDS capable of multiclass temporal analysis. Simultaneously, an IoT network environment is simulated using graph-theoretic principles, where each node represents a device characterized by parameters such as latency, energy usage, and communication protocols. Cyberattack scenarios are emulated within this network to facilitate real-time detection of anomalous behaviors. Experimental results demonstrate the effectiveness of the proposed model in capturing sequential patterns and improving detection accuracy in complex IoT environments.