Deep Learning-Based Authentication Framework for Secure Terrestrial Communications in Next Generation Heterogeneous Networks

Abstract

Smart Grid (SG) network is growing rapidly to improve the efficiency and reliability of the power supply. It will connect billions of devices to the internet in the near future. This increase in the number of internet-connected devices has led to an exponential increase in the market size of smart meters. These smart meters and devices will generate lakhs of gigabytes of data in upcoming years. To handle this massive amount of data flow, the Information and Communication Technology (ICT) of SG will experience huge advancements in terms of 5G and 6G wireless networks. These networks will provide higher data rates by utilizing the huge bandwidth spectrum of heterogeneous networks (HetNet). As the massive amount of data will be transmitted to various components of the smart grid using the internet, an open channel. Therefore, in addition to the need for high bandwidth and low latency, HetNet security has also become a prime concern. This paper presents an efficient approach to authenticate the smart devices in SG environment by using the channel state information (CSI) of the wireless links. Firstly, Low-Density Parity-Check (LDPC) codes are used to maintain the confidentiality and reliability of data transmission in the Advanced Metering Infrastructure (AMI) of SG. Then, a Convolution Neural Network (CNN)-based authentication framework is proposed to enhance the physical layer security (PLS) of terrestrial communications in next generation SG Hetnets. Further, the proposed framework is evaluated using received Channel State Information (CSI) values from smart meters placed at different locations, while also taking the scenario of active eavesdroppers into account. Experimental analysis reveals that the proposed CNN based authentication achieves 30.76 percent better authentication rate than existing DNN based authentication scheme. Further, the LDPC coded data transmission scheme reduces the bit error rate (BER) by 20 percent as compared to uncoded data transmission.