Securing Networks with Deep Learning: A Hybrid Approach to Intrusion Detection

Abstract

Traditional intrusion detection systems (IDS) are having a harder time reacting effectively as the number and complexity of cyberthreats keep rising. Presenting a novel hybrid approach to network intrusion detection that combines deep learning with conventional machine learning approaches is the aim of this study. Our suggested approach combines a random forest classifier for intrusion detection and classification with a convolutional neural network (CNN) for feature extraction from unprocessed network traffic data. This makes it possible to detect and classify incursions with the highest level of accuracy. Through the use of the NSL-KDD dataset, we compare our hybrid model to both independent deep learning and conventional machine learning methods. In contrast to single models, our CNN-Random Forest hybrid model exhibits reduced false positive rates and greater accuracy (99.2%). We also look into how computationally efficient the model is and how vulnerable it is to zero-day attacks. This study offers support for current efforts to enhance network security through the use of incredibly potent artificial intelligence-powered intrusion detection systems.