Deceptive Approach for Internet of Things Security with Machine Learning
Volviane Saphir Mfogo, Alain B. Zemkoho, Laurent Njilla, Marcellin Nkenlifack, Charles Kamhoua · 2024
In Internet of Things (IoT) security, rapid detection and mitigation of vulnerabilities are crucial to prevent unauthorized access by cyber attackers. Traditional methods such as patching often fall short due to delays in deployment and updates. This paper introduces a novel security framework that leverages machine learning (ML) to predict the vulnerabilities in IoT devices most likely to be exploited, using historical data from the Vulnerability and Attack Repository for IoT (VARIoT). Our approach employs Support Vector Machines (SVM), to identify patterns of past exploits and anticipate potential future attacks. Once vulnerabilities are predicted, our system generates a fake device profile that includes real and fabricated vulnerabilities. This profile is designed to be visible to attackers scanning the network, redirecting them to controlled environments such as honeypots when they attempt to exploit these vulnerabilities. This deceptive strategy misleads attackers and allows for the monitoring and analysis of their behavior in a safe manner, protecting the devices. Initial results demonstrate the system’s effectiveness in reducing unauthorized access and misleading potential attackers, highlighting significant advancements in proactive IoT cybersecurity.