Deeper Fine-Tuned Autoencoder for User Datagram Protocol Flooding Network Traffic Detection in Internet of Things

Abstract

Abstract The increase in the use of Internet of Things (IOT) devices operating online has led to an increase in cyber-attacks with these devices. One of the uncontrolled attacks carried out with a botnet is User Datagram Protocol (UDP) flooding. It is necessary to develop an effective method to detect abnormal UDP flooding traffic IOT devices that are compromised the IOT devices. Detection of anomalies in network traffic is the most effective method. Although machine learning, shallow neural networks and deep learning methods are used to detect abnormal traffic, in this study, it is suggested that the effective measurement metrics should be selected and applied to a fine-tuned auto-coder architecture. The main contribution of the proposed method is that a classification with high accuracy and performance can be performed by encoding the selected features deeper. The proposed method is verified with UDP-flood data in the N-BaIoT and NSL-KDD test datasets. The proposed method proved to be successful in terms of Cohen kappa, f1 score, sensitivity and accuracy metrics obtained in the experimental results. Experiments in the study showed that the number of optimally selected features was significantly reduced, resulting in the lowest detection time. This enabled a more optimized and feasible design.