Machine Learning-Based Internet of Things Attack Detection Model Over IPv6 Network

Abstract

The Internet of Things (IoT) has brought about significant advancements in smart home technology, enabling seamless communication and control of various devices. However, this increased accessibility through the internet also exposes these devices to security threats, particularly flooding attacks that can overwhelm and harm the system. To mitigate these risks, two studies propose innovative solutions. The first study introduces a scalable network capturing model using multiple Raspberry Pi boards to monitor network traffic concurrently. Evaluations demonstrate its superiority, with 30.44% more memory consumption, 14.66% lower CPU usage, and 3.63% faster execution time compared to a single capture model. The second study focuses on the vulnerabilities in IPv6 networks and designs an IPv6 flood attack detection system using the Epsilon Greedy optimized Q Learning algorithm. The proposed agent achieves 98% accuracy and 11,550 rewards, outperforming other agents, and utilizing more than 99% of a single CPU. In conclusion, these studies provide effective measures to enhance IoT network security. The scalable model and Q Learning-based detection system offer improved performance and accuracy, safeguarding IoT devices against potential threats in the connected world.