A Novel Hybrid Approach For Network Intrusion Detection Using Extreme Gradient Boosting And Long Short-Term Memory Networks
Rashmi · i-manager s Journal on Computer Science · 2021
Network Intrusion Detection (NID) has become a prominent topic nowadays with the increased use of technology and networks. In this study, a novel hybrid approach for network intrusion detection has been presented using Extreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM) networks. The benchmark NSL-KDD dataset has been used. A number of minimal feature sets were created using XGBoost for feature selection and the effects of using them in an LSTM model for detecting whether or not the network features belong to an attack were studied. It has been observed that XGBoost feature selection could be used to create minimal feature sets with very high feature reduction ratios to use in an LSTM model for NID in order to have a clear understanding of the features the model uses to learn, to achieve shorter training times and a good accuracy value close to that achieved using all the features in the dataset utilizing lower space. The findings of this study can be used for building better NID systems using deep neural networks for real-time NID. Also, they can be utilized to develop a first layer of defense for alerting the users about possible threats in real-time.