Innovative Machine Learning Strategies for Enhancing Cybersecurity Resilience in IoT Environments.

Abstract

This research investigates the incorporation of machine learning ML models in improving cybersecurity resilience for the Internet of Things (IoT) ecosystem. Since cyberattacks against IoT are on the rise, this paper investigates the efficacy of ML algorithms such as Decision Trees, Random Forests, and K-Means Clustering on common IoT attacks, DDoS (Distributed Denial of Service), spoofing, and data injection. The research builds these models on a simulated set-up with the help of widely accessible data sets and modeling tools such as Node-RED and NS3 and then validates them to check their detection rates, false positive rates, and performance in terms of system performance under such attack scenarios. It shows very high detection rates, especially for DDoS attacks (95%) and very low false positives (3%-5%). It was found that DDoS attacks had the highest increase in system latency compared to other attacks while spoofing and data injection also contributed to increasing latency but to a lesser extent. The results underscore the promising role of ML in enhancing IoT security and emphasize the need for frequent model updates and fine-tuning to address dynamic cyber risks in real-time situations. It provides a comprehensive analysis and insights into the effective use of ML models for real-time IoT security and to formulate an efficient approach for scalable IoT security solutions.