Scholarlink

<p>Nodes in a wireless sensor network (WSN) are usually contrived by hardware and environmental circumstances and high system security susceptibilities. This certainty necessitates distinctive prerequisites for designing unique network protocol, security evaluation prototypes and energy efficient techniques. Data aggregation is an efficient energy saving mechanism that eliminates unnecessary information from the aggregated data, therefore minimizing the network energy consumption significantly. However, with the deployment of sensor nodes in open environment are hence subjected to different types of attacks ushered by malicious sensor nodes. As a consequence, data aggregation leads the ways for new confrontations to WSN security. In this work a method called Federated Stochastic Gradient Averaging Ring Homomorphism-based Learning (FSGARH-L) for secure data aggregation in WSN is proposed. The FSGARH-L method is split into three sections. First, the copies of the model or data packets are shared between sensor nodes for performing training using the Federated Averaging learning model. Following which fine tune weights are generated using Stochastic Gradient Averaging function. Finally, a Federated Learning environment with heterogeneous sensor nodes positioned across the network is considered for performing secure data aggregation. The processing power of each sensor node and size of training data packets can be heterogeneous. Owing to this reason, each sensor node possesses different execution time for encoding and the response time for decoding. Higher correlated sensor nodes are obtained using the Cosine Similarity function. Following which secure data aggregation is performed using Ring Homomorphism-based encryption/decryption. The proposed FSGARH-L method is analyzed theoretically and simulated from several viewpoints. Comparisons are made with conventional methods. As a result of this study, an inclusive security solution is provided and more successful results were obtained by improving packet delivery ratio by 12%, minimizing packet drop rate and computational overhead by 38%, transmission delay by 44% and improved throughput by 37%.</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

Federated stochastic gradient averaging ring homomorphism based learning for secure data aggregation in WSN

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年5月28日

查看全文
  • Saravanakumar Pichumani,
  • T. V. P. Sundararajan,
  • S. M. Ramesh

摘要

Nodes in a wireless sensor network (WSN) are usually contrived by hardware and environmental circumstances and high system security susceptibilities. This certainty necessitates distinctive prerequisites for designing unique network protocol, security evaluation prototypes and energy efficient techniques. Data aggregation is an efficient energy saving mechanism that eliminates unnecessary information from the aggregated data, therefore minimizing the network energy consumption significantly. However, with the deployment of sensor nodes in open environment are hence subjected to different types of attacks ushered by malicious sensor nodes. As a consequence, data aggregation leads the ways for new confrontations to WSN security. In this work a method called Federated Stochastic Gradient Averaging Ring Homomorphism-based Learning (FSGARH-L) for secure data aggregation in WSN is proposed. The FSGARH-L method is split into three sections. First, the copies of the model or data packets are shared between sensor nodes for performing training using the Federated Averaging learning model. Following which fine tune weights are generated using Stochastic Gradient Averaging function. Finally, a Federated Learning environment with heterogeneous sensor nodes positioned across the network is considered for performing secure data aggregation. The processing power of each sensor node and size of training data packets can be heterogeneous. Owing to this reason, each sensor node possesses different execution time for encoding and the response time for decoding. Higher correlated sensor nodes are obtained using the Cosine Similarity function. Following which secure data aggregation is performed using Ring Homomorphism-based encryption/decryption. The proposed FSGARH-L method is analyzed theoretically and simulated from several viewpoints. Comparisons are made with conventional methods. As a result of this study, an inclusive security solution is provided and more successful results were obtained by improving packet delivery ratio by 12%, minimizing packet drop rate and computational overhead by 38%, transmission delay by 44% and improved throughput by 37%.