An Efficient Intrusion Detection Model Based on Recurrent Neural Network

Intrusion detection has proven to be an efficient strategy of information security since it can identify unknown attacks from network traffic. The existing method to detect network anomalies is often based on classic machine learning models like KNN, SVM, and others. Even though these techniques can provide some impressive results, they have a poor level of accuracy and rely primarily on manual system design is required in feature extraction which is no longer relevant in the big data era. A deep learning-based intrusion detection methodology is suggested in this method to address the issues of low accuracy and feature extraction. The recurrent neural network is used in this approach with three steps in preprocessing such as data numerical conversion, data normalization, and data balancing. It can efficiently represent network traffic flow and enhance the capacity to identify anomalies. The suggested model is put to the test using a publicly available benchmark dataset, and the findings show that it outperforms alternative comparison approaches. The result analysis of the proposed method shows that the average accuracy is 99.56%, average TPR is 99.55%, average TNR is 99.32%.