On Hybrid Artificial Neural Networks and Variational Quantum Classifier for Network Intrusion Detection

The exponential growth of internet reliance for everyday activities has opened the window for cyber-attacks deadlier. It is estimated that approx. 6.3 trillion intrusion attacks happened in 2022 alone. Thus, it is very important to utilize network intrusion detection systems (NIDS) to protect these systems. In our research, we implemented both classical and quantum algorithms for intrusion detection with real Kaggle datasets. A hybrid artificial neural network was created with 1 layer of CNN for feature extraction and an LSTM RNN model for sequential data analysis and 3-layer DNN for classification and error correction. Our classical method achieved an accuracy of around 80% sweeping all attack types. In addition, QSVM and VQC (Variational Quantum classifier) were implemented to display the potential of using quantum algorithms as standalone systems to detect network intrusion for large classical data centers. Our method achieved an accuracy of 60% with very minimal features. The desired network features and topologies are extracted from the KDDCup 99 dataset in combination with various data transformation techniques for effective classification of network intrusions and zero-day attack predictions.