IoT Network Security: NetFlow Traffic Analysis and Attack Classification Using Machine Learning Techniques

In the modern day, Internet of Things (IoT) has become critical since it permits seamless integration of physical objects and data-driven communication. IoT improves efficiency, automated processes, and real-time decision-making across the different sectors such as smart cities, transportation, agriculture, healthcare, etc. In IoT, machine learning is essential for advancing security, flexibility, and efficiency. Its real-time anomaly detection improves security by recognizing threats quickly. In this paper, we employed an IoT dataset called NF-ToN-IoT, a NetFlow-based dataset which has nine attack categories. On this dataset, we used SMOTE-ENN, a potent machine learning balancing technique created to solve the issue of class imbalance which provides best results after implementing Machine Learning(ML) models. Preprocessing was done on the dataset to eliminate extraneous or missing values. After that we applied four ML algorithms to classify attacks in NF-ToN-IoT dataset & these are Linear Discriminant Analysis (LDA), Decision Tree (DT), Random Forest (RF), and K-Nearest Neighbors (KNN) classifiers. These algorithms are quite good at identifying patterns and abnormalities in the dataset, which improves the efficiency of attack classification. The framework which we proposed achieved an accuracy rate of 60.6%, 98%, 98.3%, and 99% for LDA, RF, DT, and KNN respectively. Hence, we get the best results from KNN to classify those attacks present in NF-ToN-IoT dataset.