Network Security with Network Intrusion Detection System using Machine Learning Deployed in a Cloud Infrastructure

A software/device which monitors the network or the internet for any cyber-attack or malicious activity is called Network Intrusion Detection System (NIDS). The conventional detection systems are limited to slow detection and have to be monitored by a network engineer. This research study has attempted to improve the conventional detection systems by including powerful machine learning algorithm to the software thereby increasing its accuracy and preventing the need of a network engineer. The proposed study has used NSLKDD dataset, which has many independent features for different cyber-attacks. The data is pre-processed in order to make the data fit for the algorithm. The most important features are selected using feature selection. The model is trained using different machine learning algorithms. The accuracies of these algorithms are evaluated, and the validation metrics are calculated for the models. The trained model with best accuracy is containerized using Docker to make it machine/operating-system independent. This containerized model is stored in Docker repository. It is deployed into a cloud service provider which has many machine learning dependencies assisting the detection system. In the cloud, a virtual machine is created with required memory and processing unit. The detection system which is in Docker repository as a container is deployed into the cloud. After successful deployment, the user is provided with an Application Programming Interface (API). The user can connect to the internet and use the Network Intrusion Detection System via API to detect any attacks present in the network and secure the devices/software in the network.