Comprehensive Survey on Effective and Diverse Attack Detection Techniques in IoT

Abstract

Introduction: The rapid proliferation of the Internet of Things (IoT) has introduced significant security susceptibilities, making IoT devices frequent targets for diverse cyberattacks.

Objectives: This survey investigates attack detection techniques in IoT, emphasizing two prominent approaches: machine learning (ML) and deep learning (DL). The ability of traditional machine learning techniques, such as decision trees, support vector machines, and k-nearest neighbors, to detect anomalies and classify attacks is examined, proving its applicability to fundamental IoT security issues.

Methods: In contrast, DL approaches, includes convolutional neural networks (CNNs), recurrent neural networks (RNNs), and autoencoders, are examined for their advanced capability to automatically extract features and recognize complex attack patterns.

Results: This survey compares the strengths and limits of these approaches across diverse IoT attack scenarios, including DDoS, malware, and spoofing attacks. A particular focus is given to evaluating supervised and unsupervised learning methodologies in real-world IoT environments.

Conclusions: In order to create reliable, scalable, and adaptable threat detection systems, the survey synthesizes insights from the body of current research to identify important trends, issues, and future directions in IoT security. The results offer a thorough grasp of how ML and DL may be used to increase IoT networks' resistance to changing cyberthreats.