基于神经网络的IEC60870-5-104协议安全性研究  

Research on the Security of IEC60870-5-104 Protocol Based on Neural Networks

作　　者：丁士明 陈飞 叶华 董诗焘 DING Shiming;CHEN Fei;YE Hua;DONG Shitao(Yunnan Power Grid Co.,Ltd,Kunming Yunnan 650000)

机构地区：[1]云南电网有限责任公司,云南昆明650000

出　　处：《长江信息通信》2023年第12期104-106,共3页Changjiang Information & Communications

摘　　要：在电网系统,设备间的通信基本上基于IEC60870-5-104协议进行信息交互的,由于IEC60870-5-104协议是一个公开的报文协议且端口固定,任何设备都可以伪造相应的协议报文进行发送,容易造成网络拥堵。为了减轻网络负担,文章首次采用深度神经网络的方式识别网络中不正确的IEC60870-5-104协议报文。通过实验表明通过深度学习的方式可以有效识别不正确的报文格式正确率为97%以上,具备一定的实用性,为其它相关研究提供了借鉴意义。In the power grid distribution system,communication between devices is basically based on the IEC60870-5-104 protocol for information exchange,due to the fact that the IEC60870-5-104 protocol is a public message protocol with fixed ports,any device can forge corresponding protocol messages for sending.In order to reduce network burden,the article uses deep neural networks for the first time to identify incorrect IEC60870-5-104 protocol messages in the network.Through experiments,it has been shown that deep learning can effectively identify incorrect message formats with an accuracy rate of over 97%,which has certain practicality and provides reference significance for other related research.

关 键 词：配电网 IEC60870-5-104 深度神经网络 

分 类 号：TM73[电气工程—电力系统及自动化]