一种电力线网络上防御潜在攻击的鲁棒性消息传递机制  

A Robust Message Passing Mechanism against Potential Attacks in Power Line Networks

作　　者：王正用 白桦 吴笛 褚如旭 王韬樾 林斌 WANG Zhengyong;BAI Hua;WU Di;CHU Ruxu;WANG Taoyue;LIN Bin(Zhejiang Huayun Power Engineering Design Consulting Co.,Ltd.,Hangzhou 310004,China;Hangzhou Zhiwei Yilian Power Technology Co.,Ltd.,Hangzhou 310030,China)

机构地区：[1]浙江华云电力工程设计咨询有限公司,杭州310004 [2]杭州智微易联电力科技有限公司,杭州310030

出　　处：《电讯技术》2023年第4期556-562,共7页Telecommunication Engineering

摘　　要：网络结构上的信息传递(Message Passing,MP)机制是理解大型网络结构,实现网络节点属性预测的重要工具。然而,在电力线通信网络中,无法在终端和链路上部署复杂的身份识别验证的计算模块,导致无法建立可信的电力线通信网络MP机制,从而无法杜绝潜在攻击者的恶意对抗攻击行为,例如恶意更改电表数据、恶意隐藏终端故障等。为此,以电力节点匿名为动机,将电力节点位置隐藏至高斯噪声中,从而使攻击者无法定位具体的目标节点,抵制潜在的攻击行为。首先证明了节点定位定理,给出了图信息在傅里叶域中传递的数学模型,从而展示主流MP模型(图神经网络)定位节点的具体量化指标;然后设计了一个从高斯噪声中生成节点位置的生成器,实现节点匿名;最后设计了一个对抗生成网络,用来实现精确的电力线通信网络建模。实验表明,即使完全隐藏电力线中节点的位置,通过端到端训练,依然可以实现精确的MP,从而构建更加可信的电力线通信网络理解机制。Message passing(MP)mechanism on network structure is an important tool to understand large-scale network structure and realize attribute prediction of network nodes.However,in the power line communication(PLC)network,it is impossible to deploy complex identification and verification computing modules on the terminal and the link,resulting in the failure to build a trusted MP mechanism in the PLC network,so that the malicious counter attack behavior for potential attackers,such as malicious change of meter data,malicious concealment of terminal faults,etc.,can not be eliminated.For this reason,this paper takes the power node anonymity as the motivation to hide the power node location in Gaussian noise,so that the attacker can not locate the specific target node and resist potential attacks.Firstly,the node localization theorem is proved,and the mathematical model of graph information transfer in the Fourier domain is given,so as to show the specific quantitative index of the mainstream MP model(graph neural network).Then,a generator is designed to generate node positions from Gaussian noise to realize node anonymity.Finally,a countermeasure generation network is designed to realize accurate modeling of PLC network.Experiments show that even if the location of nodes in the power line is completely hidden,accurate MP can still be achieved through end-to-end training,so as to build a more reliable PLC network understanding mechanism.

关 键 词：电力线通信网络 消息传递 潜在攻击 恶意攻击 对抗攻击 对抗生成网络 

分 类 号：TN915.853[电子电信—通信与信息系统]