FEED-FORWARD INTRUSION DETECTION AND CLASSIFICATION ON A SMART GRID NETWORK

Cyberattack detection modeling on a smart grid and critical networks is rising due to the growing number of successful, sophisticated attacks to industrial controllers and critical infrastructure using False Data Injection Attack (FDIA), Distributed Denial of Service (DDoS), and Data replay attacks. Integrating Deep Learning algorithms, i.e., Artificial Neural Network (ANN), Convolutional Neural Network and Recurrent Neural Network (RNN), with Network Intrusion Detection Systems (NIDs), and Host-based Intrusion Detection Systems (HIDs) can improve the critical infrastructures' accuracy during anomaly detection, class classification, and ingress packet filtering. This paper leveraged the ANN's sequential classifier in training, testing, and accurately predicting the attack vectors held in the NSL-KDD dataset, which led to the proposal of a hybrid Multilayer Perceptron(MLP) Sequential-Feedforward Neural Network (SEQ-FFNN). NSL-KDD is a foundational cyber-attack modeling dataset that holds transmission protocols like TCP/IP, HTTP, and POP, critical to packet propagation in all networking infrastructures. SEQ-FFNN is a self-learning model that performs a Principal component analysis, Hyperparametization, Testing, Training, and Prediction of accurate outputs; during each code iteration. SEQ-FFNN detected and classified the attack packets to return improved accuracy of 98.97% for Tanh activated layers and 99.59% for the Sigmoid activated model.