A Power-efficient Framework for Software-defined IoT Ecosystem using Machine Learning

The paper presents a power-efficient approach for the IoT ecosystem in a software-defined network (SDN) using machine learning. At present, SDN prevails in networking infrastructures, intending to lead towards a more feasible and convenient way to manage with greater control. However, the resource constraint IoT ecosystems still lack promising power-saving features due to the broader sensor networks' mismanagement. In this work, we propose a power-efficient approach using SDN that utilizes machine learning to predict sensor devices' idle time and save power by putting them into sleep mode when not being used. The framework also includes a process calculation module that helps mitigate high-level calculation in the sensor device to locate a selfish node that may cause congestion. Priority-based device control and the machine learning module's addition reduce unnecessary power consumption while a centralized SDN controller supervises the whole topology.