Convolutional Neural Network Based Model for Intrusion Detection

Abstract

In recent research works, Machine learning techniques have been effective in identifying vulnerabilities and attacks better than most traditional methods.In this Study an Intrusion Detection Model is developed using Convolutional Neural Network (CNN) for the attack features of CICIDS-2017 dataset.CNN is excellent in Computer vision, text, and object recognition.One major benefit of CNN over Machine learning algorithms is that Feature selection is done without the need for human intervention.The proposed system is based on feature extraction and learning as predicates for prediction.The various tools deployed for developing this model include Python programming language, Microsoft Excel, Jupyter notebook of Anaconda navigator etc. Evaluation of the Model's performance was done by comparing the accuracy of the Model with other Machine learning/Deep learning IDS Models, Experimental results showed that The Proposed Model's performance was higher than the performances of the Models it was compared with, in accuracy of 99.78%.