Evolving 5G Cyber Security Using Augmented Physics-Informed Neural Networks and EfficientNetV2 for Robust Communication

In 5G wireless communications, the usage of broadcasting channel data helps identify spoof attempts. Users and their connected devices in this concept are applied to a variety of networks with activities. There has been adequate discussion on the tasks to reduce the impact of cyber-attacks in genuinely complex networks and highlight their potential effectiveness. These are extremely susceptible to a variety of technologies at various conceptual levels, like 5G broadband networks. In this paper, an Evolving Cyber Security System entrenched on Augmented Physics-Informed Neural Networks with EfficientNetV2 for 5G wireless Communication Networks (CSS-APINN-ENetV2-5GWCN) is proposed. Initially, the data is taken from the UNSW-NB15 Dataset and pre-processed using UN sharp Structure Guided Filtering (USGF) for normalizing data. The Fractional-Order Water Flow Optimizer (FOWFO) is used for selecting the optimal features. The selected features are given to Augmented Physics-Informed Neural Networks (APINN) with EfficientNetV2 (ENetV2) (APINN-ENetV2) to classify the attacks as DOS, Normal, remote to user, User to root and Probing attack. Homonuclear Molecules Optimization Algorithm (HMOA) optimizes the weight parameters of APINN-ENetV2. The experimental outcomes of the proposed CSS-APINN-ENetV2-5GWCN approach attain 24.39%, 35.71%, and 25.55% higher G-mean; 20.73%, 13.79%, and 16.47% higher correlation coefficient; 24.44%, 35.18%, and 14.44% higher identification rate compared with the existing techniques.