A convolutional neural network based model with improved activation function and optimizer for effective intrusion detection and classification

Abstract

Technological developments in today's world have tied our financial, social and other facets of life to the Internet, expanding into our transportation, home appliances and more device with rising IoT technologies. Additionally, the dramatic growth in number of cyber-Attacks have presented our sensitive information on Internet with critical security risks. To fix this problem, the Intrusion Detection System acts as a realistic method to identify cyber-Attacks when they are underway or prior to them. Leveraging Deep Learning methods, we can utilize the most sophisticated multi-functional architectures available right now to identify and classify intrusions with maximum precision. Our paper proposes a Convolutional Neural Network model with mish activation function and Ranger optimizer that reaches a higher degree of precision compared to previous Deep Learning models and traditional CNN models that utilize ReLU activation function and Adam optimizer. In this study, we have used the relatively novel dataset CIC-IDS-2018 for testing which comprises six different varieties of attacks. Our model reaches an unprecedented accuracy of 98.9% that is the highest in multiclass classification with this dataset.