DDoS Intrusions Detection in Low Power SD-IoT Devices Leveraging Effective Machine Learning

Abstract

Security and privacy are significant concerns in software-defined networking (SDN)-applied Internet of Things (IoT) environments, due to the proliferation of connected devices and the potential for cyberattacks. Hence, robust security mechanisms need to be developed, including authentication, encryption, and distributed denial of service (DDoS) attack detection, tailored to the constraints of low-power IoT devices. Selecting a suitable tiny machine learning (TinyML) algorithm for low-power IoT devices for DDoS attack detection involves considering various factors such as computational complexity, robustness in dealing with heterogeneous data, accuracy, and the specific constraints of the target IoT device. In this paper, we present a two-fold approach for the optimal TinyML algorithm selection leveraging the hybrid analytical network process (HANP). First, we make a comparative analysis (qualitative) of the machine learning algorithm in the context of suitability for TinyML in the domain of SD-IoT devices and generate the weights of suitability for TinyML applications in SD-IoT. Then we evaluate the performance of the machine learning algorithms and validate the results of the model to demonstrate the effectiveness of the proposed method. Finally, we see the effect of dimensionality reduction with respect to features and how it affects the precision, recall, accuracy, and F1 score. The results demonstrate the effectiveness of the scheme.