Enhancing Security Using Deep Learning Based Capsule Neural Networks (CAPSNETS) with Enhanced Animal Migration Algorithm (EAMO) In Wireless Sensor Network
Year: 2025
Type: article
Abstract: The rapid development of wireless communication and sensor technologies has led to a development in the application of WSN in a variety of areas, including industry, aviation, and the environment.This development overcomes the drawbacks of conventional techniques for Data Collection (DC) and monitoring in challenging conditions [1].However, new threats are brought about by the WSN security problem.WSN intrusion attacks have the potential to seriously risk the security of the lives and property of individuals or groups.Reliable security assurance for the network can be obtained through Accurate Detection (AD) of different kinds of attacks in WSN.Consequently, one of the crucial domains in the network security nowadays is the WSN Intrusion Detection (ID) technique.A distributed wireless communication network consisting of management, sink (Snk), and SN nodes is called a WSN.Wireless signals can be divergent because SN is limited by their own computational capability, storage space, and capacity for wireless communication.Data forgeries and leaks are among the security risks that WSN faces [2].To guarantee WSN security, current research typically uses a two-layer Defence Mechanism (DM).Security protocols, data encryption, and Data Authentication (DA) make up the first layer of the DM.However, as network attack technologies continue to advance, the effectiveness of the first line of defence increasingly deteriorates.In order to mitigate the drawbacks of the initial layer DM and minimize losses resulting from network attacks, ID technology serves as the second layer ABSTRACT: Wireless Sensor Networks (WSNs) are widely used in mission-critical domains such as environmental monitoring, industrial automation, and defense.However, they face two major challenges: vulnerability to malicious node attacks and high energy consumption, both of which degrade network performance and shorten operational lifespan.This study aims to develop a unified framework that simultaneously addresses security and energy efficiency in WSNs.The novelty lies in integrating Capsule Neural Networks (CapsNets) with the Enhanced Animal Migration Optimization (EAMO) algorithm-a combination not explored in existing methods.Unlike conventional intrusion detection models that rely on CNNs or shallow classifiers, CapsNets preserve spatial relationships in features through vector-based representation and dynamic routing, improving detection accuracy and reducing false positives.Simultaneously, EAMO-enhanced with birth-death processes-optimizes LEACH-based routing by selecting energy-aware cluster heads and secure paths, reducing latency and energy drain.Using Z-Score Normalization for preprocessing, Fuzzy C-Means clustering, and the NSL-KDD dataset, the framework is evaluated in NS2 simulations.Results show superior packet delivery ratio, detection accuracy, and energy efficiency compared to ASNGSRA, DMCNN, and FRCSROD protocols.This integrated deep learning-metaheuristic approach offers a robust, scalable solution for secure, energy-aware WSNs, with potential future work focusing on adaptive defenses against evolving cyber threats.
Authors: S. Vishalini, A. Kavitha
Institution: Kongunadu Arts and Science College
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7087760992
FWCI:
Cites:
Cited by:
Related to: