A Machine Learning-based SDN Controller Framework for Drone Management

Abstract

With the advancement of information and communication technology, Unmanned Aerial Vehicles (UAV), popularly known as drones, have also increased. The drones have been noted for their wide range of applications such as military, search and rescue operation, disaster detection and monitoring, agriculture, and delivery. Each type of drone has different characteristics and functionality based on its application, making them a security threat for some city zone. Therefore, there is an essential need for efficient drone management based on their type and application in different zones. To do this, we proposed a Machine learning (ML) based Software Defined Network (SDN) drone management framework. In this framework, the SDN controller uses ML with the drone’s radio frequency feature to detect its type and application and, according to its application, authenticate it and assign communication rules. SDN controller records authentication information in a DAG-based Distributed Ledger Technology (DLT) available for other SDN controllers. When a drone desires to migrate to another zone, the destination SDN controller can achieve authentication information by referring to DAG-based DLT, and there is no need for re-authentication. The experimental result shows authentication delay reduction in our proposed framework. Moreover, we adopted ML algorithms includes Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR), to evaluate our proposed framework in drone’s type classification. The result shows that the RF algorithm shows the best performance with 92.81% accuracy in the classification of the drone’s type.