Leveraging Machine Learning and SDN-Fog Infrastructure to Mitigate Flood Attacks

Abstract

The use of IoT devices is growing rapidly and it is playing a critical role in a diverse set of industries. It has been instrumental in the growth of smart cities. Smart cities have emerged as a paradigm for urban development which aims to be sustainable, efficient and improve accessibility. However, the limited processing power of IoT devices makes them susceptible to flood-based attacks. Denial of Service attacks can overwhelm the computing resources or network bandwidth of IoT networks. Since IoT devices power critical infrastructure like traffic management in smart cities, adequate defense of such networks from malicious actors is imperative. In this article, the authors propose a framework tailored for detection and mitigation of flood-based attacks in smart city infrastructure. The proposed smart city framework aims to reduce latency of attack detection by using fog computing for feature extraction and security maintenance. It allows scalability by utilizing SDN and fog infrastructure for mitigation of attacks. We have analysed and utilized packet-level features which are excellent for distinguishing between IoT and attack traffic. We have trained and quantitatively compared 5 state-of-the-art supervised machine learning models for attack detection in this paper. We were able to achieve an accuracy of 99.9% on our simulated dataset in attack detection.