A Literature Review on Integrating Hybrid Intrusion Detection Systems

The research provides a systematic review on the development of hybrid intrusion detection systems that combine signature-based and anomaly-based techniques to improve identification of attacks. The study analyzes different methods of data preprocessing, including feature selection and class imbalance mitigation, as well as model evaluation using popular datasets such as NSL-KDD, CICIDS2017, and UNSW-NB15. In addition, the integration of federated learning techniques was introduced as an approach to enhance data privacy and system scalability. This research also outlines the primary unexplored gaps in the standardization of evaluation and real-world testing of data, alongside offering tailored recommendations aimed towards developing more streamlined and efficient Intrusion Detection Systems in the future. Systematic Review Self 10 Studies highlighted the fact that the concentration was primarily focused on classification of IDS datasets. We employed machine learning techniques using Support Vector Machine, Random Forest, XGBoost, Gradient Boosting, Logistic Regression, and K-Nearest Neighbors. In deep learning there are techniques that include CNN and RNN. XGBoost and RF were remarkable in achieving accuracy on the IIS dataset but several others outperformed them on F1 score. This study illustrates the latest research in classifying datasets on IDS with various methods to determine the best one and proposes directions for future research to improve accuracy and model categorization on IDS such as Hybrid IDS and FL-IDS.