Using Ensemble Model for Classifying Network Traffic Flow and Identifying Attacks in Smart City

Objectives:Smart City generates the vast amount of data through sensors and actuators. Smart city consist of diverse data that rely on IoT which demands the effective intrusion detection system. Paper focuses on key mechanism of Quality of Service to provide the security. This study aims to provide light-weight solution to classify the network attack categories using a robust approach. Methods: To detect the exact type of attack we have employed the Multi-class classification on the NF-TON-IOT-v2 dataset. This is a critical area of research in network traffic classification, particularly for enhancing IoT security. Random Forest, KNN and Ensemble models have been applied to this dataset, demonstrating significant advancements in accuracy and efficiency. Findings: The proposed Ensemble Model approach offers efficient and accurate traffic classification to detect attacks on the network. System gives 96.45% results with Ensemble algorithm, whereas Random Forest and KNN gives the accuracy of 96.15% and 95.15% respectively. Novelty: Proposed Ensemble model were trained on only 14 features out of 43, which were identified by the hybrid feature importance method. Unlike other existing deep learning studies, our study minimized the overburden of model by reducing the dimensions while providing promising result for Multiclass Classification. Keywords: Attack Detection, NF-TON-IoT, Traffic Classification, QoS, Security Mechanism, Ensemble Model, Net-flow, Intrusion Detection System