A Hybrid Deep Reinforcement Learning based Secure Multi-Hop Data Aggregation Framework for Wireless Sensor Networks

Abstract

Wireless Sensor Networks (WSNs) require secure, scalable, and energy-efficient data aggregation mechanisms due to limited node resources, dynamic topology changes, malicious attacks, and congestion issues. Existing models such as QGAOA, RTMA², and RTCAMA² have significantly improved cluster head selection, trust-aware routing, and congestion-aware aggregation; however, these approaches remain limited by static optimization strategies and insufficient adaptability to rapidly changing network conditions. To address these challenges, this paper proposes a Hybrid Deep Reinforcement Learning-based Secure Multi-Hop Data Aggregation Framework (HDRL-SMDAF) for WSNs. The proposed framework integrates Deep Q-Network (DQN)-based adaptive cluster head selection, dynamic trust recalibration, congestion prediction, and secure multi-hop routing. Reinforcement learning enables sensor nodes to dynamically optimize routing and aggregation decisions based on residual energy, trust metrics, congestion level, packet delivery ratio, and network lifetime.

The framework enhances security by isolating malicious nodes while simultaneously optimizing energy consumption and throughput. Simulation results demonstrate that HDRL-SMDAF outperforms existing models in terms of network lifetime, packet delivery ratio, throughput, delay reduction, energy consumption, and trust resilience. This proposed model offers a scalable next-generation solution for intelligent WSN deployments in IoT, smart surveillance, healthcare, and industrial monitoring applications.