Scholarlink

<p>Wireless sensor networks (WSNs) are emerging as a vital technology for networking applications due to their wide range of applications and cost-effectiveness. Combinatorial design-based group key management provides a reliable and flexible security mechanism for secure group communication in WSNs by efficiently assigning the number of keys per node. In this work, we used a dataset of 400 designs from symmetric balanced incomplete block designs (SBIBD). The SBIBD is an arrangement of a finite set into subsets satisfying balance properties such as the set of elements being equal to the collection of blocks, the size of the block being equal to the number of repetitions of elements in a block, and every pair-wise element contains in exactly one block. The dataset was used to train four machine learning models: support vector machine, artificial neural networks, K-nearest neighbours, and extreme learning machine. We observed that among these four algorithms, the extreme learning machine demonstrated a higher accuracy in predicting re-keying values with the coefficient of determination, mean squared error, root mean squared error, and mean absolute error values, respectively.</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 Machine learning approach to predict re-key in wireless sensor network using symmetric balanced incomplete block designs

  • 期刊论文

  • 发表日期：

    2026年3月19日

查看全文
  • Prity Kumari,
  • Ranjan Kumar

摘要

Wireless sensor networks (WSNs) are emerging as a vital technology for networking applications due to their wide range of applications and cost-effectiveness. Combinatorial design-based group key management provides a reliable and flexible security mechanism for secure group communication in WSNs by efficiently assigning the number of keys per node. In this work, we used a dataset of 400 designs from symmetric balanced incomplete block designs (SBIBD). The SBIBD is an arrangement of a finite set into subsets satisfying balance properties such as the set of elements being equal to the collection of blocks, the size of the block being equal to the number of repetitions of elements in a block, and every pair-wise element contains in exactly one block. The dataset was used to train four machine learning models: support vector machine, artificial neural networks, K-nearest neighbours, and extreme learning machine. We observed that among these four algorithms, the extreme learning machine demonstrated a higher accuracy in predicting re-keying values with the coefficient of determination, mean squared error, root mean squared error, and mean absolute error values, respectively.