物联网学报2025，Vol.9Issue(2)：107-116,10.DOI:10.11959/j.issn.2096-3750.2025.00455
面向资源受限电力物联网终端的语义安全通信方法
A semantic secure communication approach for resource-constrained power IoT terminals
摘要
Abstract
The training of semantic models typically consumes a large amount of energy and time,which hinders the implementation of semantic transmission on power Internet of things(IoT)terminals with limited resources.To reduce the energy and time consumption at the terminal,a novel semantic communication architecture was proposed.Firstly,the data to be transmitted was uploaded to a machine learning as a service(MLaaS)platform.Then,the semantic model training was performed on the MLaaS platform,and the model parameters were sent back to the terminal.Finally,semantic reason-ing was carried out through the terminal.However,this architecture faces the risk of the MLaaS platform leaking seman-tic model parameters,leading to the potential eavesdropping of semantic information.Therefore,an anti-eavesdropping method based on feature obfuscation was designed to address the security communication issue of the MLaaS platform during the semantic inference phase.Experimental results show that the proposed method is effective against passive eavesdroppers.The effectiveness of the proposed anti-eavesdropping method was demonstrated.Additionally,the compu-tational overhead and latency of the data obfuscation module on terminals have been preliminarily verified,showing that the method is feasible for practical application on resource-constrained power IoT terminals.关键词
电力物联网终端/语义通信/机器学习即服务/抗窃听Key words
power IoT terminal/semantic communication/machine learning as a service/anti-eavesdropping分类
信息技术与安全科学引用本文复制引用
杨会峰,尚立,崔俊彬,刘红艳,王九成,蔺鹏..面向资源受限电力物联网终端的语义安全通信方法[J].物联网学报,2025,9(2):107-116,10.基金项目
国网河北省电力有限公司科技项目(No.SGHEXT00GCJS2310140)The Science and Technology Project of State Grid Hebei Electric Power Co.,Ltd.(No.SGHEXT00GCJS2310140) （No.SGHEXT00GCJS2310140）
