Evaluation of Network Intrusion Detection with Machine Learning and Deep Learning Using Ensemble Methods on CICIDS-2017 Dataset

An intrusion detection system (IDS), also referred to as an IDS, is a type of network security equipment that monitors network communications in real-time and, should any potentially hostile transmissions be identified, either sends out alarms or implements active reaction measures. In this context, numerous researchers have attempted to improve intrusion detection performance by combining traditional machine learning models with alternative optimization techniques. Although the present intrusion detection model has the potential to considerably improve performance, there are still certain persistent problems, such as faulty detection and data preparation operations, which both have the potential to reduce accuracy. Using the CICIDS2017 dataset, we provided an analysis model of various classifiers in our paper, including random forest, stacking set-1 (Naive Bayes, K-Nearest Neighbours, Random Forest), bagging, boosting (XGB), LSTM, CNN, and stacking set-2 (LSTM and CNN). Boosting (XGB) performed better than all models across all feature sets overall.