A Comparative Analysis of Machine Learning Algorithms for Intrusion Detection in Edge-Enabled IoT Networks

TL;DR

This paper compares various machine learning algorithms for intrusion detection in edge-enabled IoT networks, highlighting the effectiveness of Multi-Layer Perception with specific performance metrics.

Contribution

It provides a comparative analysis of ML algorithms for IoT intrusion detection, identifying MLP as a suitable choice for edge-based networks based on accuracy and training time.

Findings

MLP achieved 79% accuracy in intrusion detection.

MLP had a training time of 1.2 seconds.

The study used the NSL-KDD dataset for evaluation.

Abstract

A significant increase in the number of interconnected devices and data communication through wireless networks has given rise to various threats, risks and security concerns. Internet of Things (IoT) applications is deployed in almost every field of daily life, including sensitive environments. The edge computing paradigm has complemented IoT applications by moving the computational processing near the data sources. Among various security models, Machine Learning (ML) based intrusion detection is the most conceivable defense mechanism to combat the anomalous behavior in edge-enabled IoT networks. The ML algorithms are used to classify the network traffic into normal and malicious attacks. Intrusion detection is one of the challenging issues in the area of network security. The research community has proposed many intrusion detection systems. However, the challenges involved in selecting suitable algorithm(s) to provide security in edge-enabled IoT networks exist. In this paper, a comparative analysis of conventional machine learning classification algorithms has been performed to categorize the network traffic on NSL-KDD dataset using Jupyter on Pycharm tool. It can be observed that Multi-Layer Perception (MLP) has dependencies between input and output and relies more on network configuration for intrusion detection. Therefore, MLP can be more appropriate for edge-based IoT networks with a better training time of 1.2 seconds and testing accuracy of 79%.