A Hybrid Intrusion Detection System Using Machine Learning and Deep Learning

Intrusion detection systems (IDS) play a significant role in ensuring the protection of networks against emergent threats since they enable the detection of intrusive events and other irregularities occurring in network traffic. The existing signaturebased and anomaly-based IDS systems have weaknesses in detecting zero-day attacks, reducing false-positive rates, and dealing with big data sets. Based on these difficulties, this study proposes a hybrid IDS framework based on machine learning and deep learning algorithms. The system composes a Random Forest classifier for the primary filtration of anomalies with an LSTM deep learning model for identifying intricate patterns of attacks. For training and evaluation, the CICIDS2017 dataset, which covers numerous varieties of attack types, was used. The experimental outcome shows that the proposed hybrid IDS yields better results of accuracy 97.8 % and precision 97.4 % over baseline and state-of-art models with recall <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 6. 5 \%}$</tex> and ROC-AUC <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 8. 5 \%}$</tex>. Moreover, it is impressive that the system also has low false favorable rates and can develop a policy for dynamically changing networks. There are three critical assumptions for future work: real-time deployment, better modelling for imbalanced data, and better robustness against adversarial attacks. The research presented above proves the effectiveness of hybrids in enhancing the possibilities of intrusion detection and network security.