【Author】
Wadhwa, Shivani; Rani, Shalli; Kaur, Gagandeep; Koundal, Deepika; Zaguia, Atef; Enbeyle, Wegayehu
【Source】WIRELESS COMMUNICATIONS & MOBILE COMPUTING
【Abstract】Cognitive learning is progressively prospering in the field of Internet of Things (IoT). With the advancement in IoT, data generation rate has also increased, whereas issues like performance, attacks on the data, security of the data, and inadequate data resources are yet to be resolved. Recent studies are mostly focusing on the security of the data which can be handled by blockchain. Blockchain technology records the learned data into the block which is generated after completing proper consensus mechanism. In this paper, Hetero Federated Learning approach is used to apply cognitive learning on data produced by Internet of Thing devices. Security on cognitiveIoT data is provided by blockchain using Proof of Work consensus mechanism. By applying blockchain over heteroFL approach, we have conducted various simulations to check the performance of our proposed framework. Parameters taken into consideration during performance evaluation are effect of number of blocks on memory utilization and impact of data sample size on accuracy according to different learning rates.
【标题】基于 HeteroFL 区块链方法的认知物联网安全
【摘要】认知学习在物联网 (IoT) 领域逐渐繁荣。随着物联网的发展，数据生成率也随之提高，而性能、数据攻击、数据安全、数据资源不足等问题仍有待解决。最近的研究主要集中在区块链可以处理的数据的安全性上。区块链技术将学习到的数据记录在完成适当的共识机制后生成的块中。在本文中，异构联邦学习方法用于将认知学习应用于物联网设备产生的数据。认知物联网数据的安全性由区块链使用工作证明共识机制提供。通过在heteroFL方法上应用区块链，我们进行了各种模拟来检查我们提出的框架的性能。在性能评估过程中考虑的参数是块数对内存利用率的影响以及数据样本大小对不同学习率的准确性的影响。
评论