Convolutional Neural Networks with DeepInsight for Image-Based Intrusion Detection

Information and network security remain critical challenges for most organizations since cyberattacks and security vulnerabilities are increasing in regularity and sophistication. With an increase in the number, intensity and variety of these attacks, traditional firewalls and data encryption methods can no longer meet the needs of current network security. In this paper, a Convolutional Neural Network (CNN) based Intrusion Detection System (IDS) was proposed to detect emerging attacks. The CSE CIC-IDS-2017 dataset was used for the development, incorporating feature engineering techniques, treating class imbalance and hyper parameter tuning for optimal results. Experimental results of CNN based NIDS model gave a 95.2% accuracy, 89.7% precision, 87.3% recall and 88.3% F1-Score.