Enhanced network connectivity with deep learning enabled intrusion detection–based security mechanism for cluster based internet of things

The rapid expansion of Internet of Things (IoT) networks has introduced important challenges related to network connectivity, energy efficiency, and security due to large-scale deployment, node heterogeneity, and resource constraints. Existing methods often address connectivity and intrusion detection independently, leading to decreased network reliability and improved vulnerability to attacks. To overcome these limitations, this study aims to improve an integrated framework that simultaneously improves connectivity, optimizes energy usage, and ensures robust intrusion detection in cluster-based IoT environments. In this work, an Enhanced Network Connectivity with Deep Learning–enabled Privacy-Preserving Mechanism (ENC-DLPPM) is proposed, incorporating Artificial Fish Swarm-based Energy-Efficient Clustering (AFS-EEC), Deep Q-Network (DQN)-based adaptive connectivity, and an Optimal Deep Canonically Correlated Sparse Autoencoder (ODCCSAE) for intrusion detection. The AFS-EEC technique selects optimal cluster heads based on energy, distance, and delay, while the DQN dynamically learns routing strategies to maintain stable connectivity and avoid isolated nodes. Additionally, the ODCCSAE model, optimized utilizing quasi-oppositional Salp Swarm Optimization (QOSSO), detects malicious activities utilizing locally extracted features, thereby improving security without exposing sensitive data. Experimental results establish that the proposed method accomplishes greater performance, with a network lifetime of up to 5,785 rounds, reduced energy consumption (45 mJ), high throughput (69.05 Kbps), minimal delay (2.18 s), and an intrusion detection accuracy of 99.83%. These findings confirm that the ENC-DLPPM framework provides an effective, scalable, and intelligent solution for achieving reliable connectivity and secure communication in next-generation IoT networks.