Scholarlink

Purpose <p>Wireless Sensor Networks (WSNs) are widely deployed in critical applications such as healthcare monitoring, automotive systems, and emergency response. However, their performance is constrained by limited energy resources, restricted computational capabilities, dynamic network conditions, and vulnerability to Denial of Service (DoS) attacks. This research aims to develop a robust, energy-efficient, and secure multi-objective framework for effective DoS attack detection and prevention in WSNs.</p> Methods <p>The proposed framework have four main components. The proposed approach uses a combination of two algorithms: First, Low Energy Adaptive Clustering Hierarchy Genetic Algorithm with Kookaburra Optimization Algorithm (LEACHGA-KOA), which improves energy usage, and extends network lifetime. Second, an improved red piranha optimization algorithm is applied to optimize feature selection, which includes an adaptive search process and a differential security mechanism to lower the feature dimension while keeping data privacy. Third, a Bayesian defense soft-swish physics-guided linear scaling neural network is developed to accurately and robustly classify by introducing Bayesian inference for uncertainty modeling, Soft-Swish activation for capturing complex non-linear patterns, and physics-guided scaling to reduce false positives under dynamic conditions. Finally, an optimized threshold function-based puzzle generator is introduced as a dynamic prevention mechanism to attack DoS attacks by adaptive puzzle generation. The components all fit into a single logical framework to allow optimization in the clustering, feature selection, classification and prevention processes to be done in a coordinated way.</p> Results <p>Experimental results demonstrate superior performance, achieving 98% accuracy, 98% precision, 97% recall, and 98% F1-score, along with improved energy efficiency and reduced computational complexity.</p> Conclusion <p>The proposed integrated framework provides a scalable, energy-aware, and highly secure solution for WSN environments, ensuring reliable operation under dynamic and adversarial conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimized Multi-Objective Cluster-Based Intrusion Detection with Bayesian Soft-Swish Neural Networks and Puzzle-Based Prevention for Denial-of-Service Attacks in Wireless Sensor Networks

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • N. Sumalatha,
  • K. Baskaran

摘要

Purpose

Wireless Sensor Networks (WSNs) are widely deployed in critical applications such as healthcare monitoring, automotive systems, and emergency response. However, their performance is constrained by limited energy resources, restricted computational capabilities, dynamic network conditions, and vulnerability to Denial of Service (DoS) attacks. This research aims to develop a robust, energy-efficient, and secure multi-objective framework for effective DoS attack detection and prevention in WSNs.

Methods

The proposed framework have four main components. The proposed approach uses a combination of two algorithms: First, Low Energy Adaptive Clustering Hierarchy Genetic Algorithm with Kookaburra Optimization Algorithm (LEACHGA-KOA), which improves energy usage, and extends network lifetime. Second, an improved red piranha optimization algorithm is applied to optimize feature selection, which includes an adaptive search process and a differential security mechanism to lower the feature dimension while keeping data privacy. Third, a Bayesian defense soft-swish physics-guided linear scaling neural network is developed to accurately and robustly classify by introducing Bayesian inference for uncertainty modeling, Soft-Swish activation for capturing complex non-linear patterns, and physics-guided scaling to reduce false positives under dynamic conditions. Finally, an optimized threshold function-based puzzle generator is introduced as a dynamic prevention mechanism to attack DoS attacks by adaptive puzzle generation. The components all fit into a single logical framework to allow optimization in the clustering, feature selection, classification and prevention processes to be done in a coordinated way.

Results

Experimental results demonstrate superior performance, achieving 98% accuracy, 98% precision, 97% recall, and 98% F1-score, along with improved energy efficiency and reduced computational complexity.

Conclusion

The proposed integrated framework provides a scalable, energy-aware, and highly secure solution for WSN environments, ensuring reliable operation under dynamic and adversarial conditions.