A Novel Hybrid Deep Learning Approach for Cyber Security Intrusion Detection

Abstract

This paper addresses the new challenge of develop a novel hybrid ANN model that will be effective in detecting the real time network anomalies. The entire system will be designed, developed, and evaluated within this paper. The main goal is to develop a Artificial neural network based algorithm and use the hyperparameter tuning features to fine tune the model and get the best model as the output to use in in a flask application. A live network capturing module will be implemented and the flask application is designed to get the lively captured data into the ANN model perform the predictions and passes to the Web application to showcase to the end user. To train the data model, a cleaned, preprocessed dataset was used to achieve the higher accuracy, performance, and efficiency rates. The experimental results demonstrate the capability of the end Intrusion detection system to identify the true network attacks as network anomaly within a very short time. The author explored the integration of RFE mechanism to use the feature selection for the model training as a extra achievement as well. With the increasing volume of the network attacks started against the complex networks within the cybersecurity industry, author wanted to deliver this system and address the issue even for some distance. At the end comprehensive evaluations were carried out by the experts as well as the author to prove the capability of the proposed system.