Scholarlink

<p>Attacks on cognitive radio networks (CRN) during spectrum sensing could severely degrade their performance. Among the many proposed solutions to mitigate the effects of these attacks, cutting-edge technologies such as machine learning and deep learning have emerged as promising tools for discovering and repairing network vulnerabilities. Improving CRN's performance and dependability is the main goal of this project, which aims to address these difficulties. In order to improve the data quality and remove outliers, preprocessing was carried out using Tukey's rule on a Kaggle dataset consisting of spectrum sensing findings. System makes use of energy detection, which is a cheap way to sense things without having to know who main users are in advance. In order to arrive at&#xa0;overall choice,&#xa0;fusion center uses&#xa0;“AND” rule to combine local judgments made by secondary users, who compare observed signal energy having threshold. In order to determine if a user is trustworthy or malevolent, the Swin Transformer (EffSwinNet) analyses these local judgments. The suggested method uses a refined version of the EffSwinNet model, which is well-known for its proficiency in dealing with hierarchical feature representations, to classify CRN. A new Lemurs Optimization technique is presented to optimize the model's hyperparameters, guaranteeing peak performance. An impressive testing accuracy of 99.88% is achieved by&#xa0;suggested procedure. Using the improved EffSwinNet model in tandem with Lemurs Optimization demonstrates as this method can distinguish between legal and illegal CRN users, which is product's unique selling point.</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

Optimal threat detection in cognitive radio networks using optimized effSwinNet and lemurs algorithm

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Suriya Murugan

摘要

Attacks on cognitive radio networks (CRN) during spectrum sensing could severely degrade their performance. Among the many proposed solutions to mitigate the effects of these attacks, cutting-edge technologies such as machine learning and deep learning have emerged as promising tools for discovering and repairing network vulnerabilities. Improving CRN's performance and dependability is the main goal of this project, which aims to address these difficulties. In order to improve the data quality and remove outliers, preprocessing was carried out using Tukey's rule on a Kaggle dataset consisting of spectrum sensing findings. System makes use of energy detection, which is a cheap way to sense things without having to know who main users are in advance. In order to arrive at overall choice, fusion center uses “AND” rule to combine local judgments made by secondary users, who compare observed signal energy having threshold. In order to determine if a user is trustworthy or malevolent, the Swin Transformer (EffSwinNet) analyses these local judgments. The suggested method uses a refined version of the EffSwinNet model, which is well-known for its proficiency in dealing with hierarchical feature representations, to classify CRN. A new Lemurs Optimization technique is presented to optimize the model's hyperparameters, guaranteeing peak performance. An impressive testing accuracy of 99.88% is achieved by suggested procedure. Using the improved EffSwinNet model in tandem with Lemurs Optimization demonstrates as this method can distinguish between legal and illegal CRN users, which is product's unique selling point.