Scholarlink

<p>When a Wireless Sensor Network successfully executes a transaction within a finite timeframe and with minimal resource consumption, it is deemed active and operational. The underlying challenge in all scenarios boils down to a common issue as Denial of Service. This challenge encompasses endeavours to enhance the enforcement of consumption and routing policies, striving for improved security and reliability. In this study, we leverage a machine learning approach, utilizing a covariance vector derived from eigenvalues to identify anomalous behaviour in nodes. The CDS parameter of each node proves pivotal in adapting the selection policy to the current conditions, guiding a suitable course of action. Decision values for likelihoods are dynamically adjusted, serving as either rewards or punishments based on prior actions. A real-time payment function is established as a reward mechanism, contingent on the action taken and its utility, such as whether a forwarded packet was dropped or delayed. At the second level, the same learning model is employed to discern between malicious and benign nodes. The proposed method showcases a notable enhancement, including a 98% improvement in Packet Delivery Ratio (%), 84% reduction in delay (seconds), 93.2% decrease in jitter (seconds), and an 86% boost in Goodput (Kilobits per second). These substantial improvements validate the efficacy of the proposed method when compared to existing methodologies employed in the same research context.</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

A Crossbreed and Powerful Machine Learning Archetype for Unparalleled Safety and Effectiveness in Wireless Sensor Networks

  • 期刊论文

  • 发表日期：

    2025年3月25日

查看全文
  • Deepak Sharma,
  • K. V. Shahnaz,
  • T. Helan Vidhya,
  • Korra Srinivas,
  • Lakshmi Ramani Burra,
  • Yeruva Jaipal Reddy,
  • Shubham Joshi

摘要

When a Wireless Sensor Network successfully executes a transaction within a finite timeframe and with minimal resource consumption, it is deemed active and operational. The underlying challenge in all scenarios boils down to a common issue as Denial of Service. This challenge encompasses endeavours to enhance the enforcement of consumption and routing policies, striving for improved security and reliability. In this study, we leverage a machine learning approach, utilizing a covariance vector derived from eigenvalues to identify anomalous behaviour in nodes. The CDS parameter of each node proves pivotal in adapting the selection policy to the current conditions, guiding a suitable course of action. Decision values for likelihoods are dynamically adjusted, serving as either rewards or punishments based on prior actions. A real-time payment function is established as a reward mechanism, contingent on the action taken and its utility, such as whether a forwarded packet was dropped or delayed. At the second level, the same learning model is employed to discern between malicious and benign nodes. The proposed method showcases a notable enhancement, including a 98% improvement in Packet Delivery Ratio (%), 84% reduction in delay (seconds), 93.2% decrease in jitter (seconds), and an 86% boost in Goodput (Kilobits per second). These substantial improvements validate the efficacy of the proposed method when compared to existing methodologies employed in the same research context.