Development of a network intrusion detection system using hybridised Recurrent Neural Networks and Random Forest models

Cyber threats are continually rising in this age because bad actors are trying to take advantage of the people, organizations, and systems. Recent advancements in artificial intelligence have released powerful agents at humanity’s fingertips. As inclusive as it is, it has created more space for potential bad actors. Systems that effectively oppose these threat actors must be developed to save humanity. In this study, hybridized models of Recurrent Neural Networks (RNN) and Random Forest classifiers are hybridised for the creation of a Network Intrusion Detection System (NIDS) using the benchmark dataset (CICIDS 2017) The necessity for an effective and precise technique to identify network intrusions, including both known and zero-day anomalies, is the main issue addressed. By using a hybrid modeling method, this study seeks to improve intrusion detection systems' accuracy and dependability. Metrics like accuracy, precision, recall, F1 score, true positive rate, and true negative rate were used to assess the developed model's performance. The hybrid model demonstrated exceptional performance with a testing accuracy of 96.08%, precision and recall of 96.0%, and an F1 score of 96.0%. The experiment outcomes demonstrate that the model performs effectively and, when deployed, can identify and categorize cyberattacks in contemporary settings.