Detection of UDP Flooding DDoS Attacks on IoT Networks Using Recurrent Neural Network

Internet of Thing (IoT) is a concept where an object can transfer data through a network without requiring human interaction. Complex IoT networks make it vulnerable to cyber attacks such as DDoS UDP Flood attacks, UDP Flood attacks can disrupt IoT devices. Therefore, this study proposes an attack detection method using a deep learning approach with the Recurrent Neural Network (RNN) method. This study uses Principle Component Analysis (PCA) to reduce the feature dimension, before learning using RNN. The purpose of this study is to test the combined performance of the PCA and RNN methods to detect DDoS UDP Flood attacks on IoT networks. The testing in this study used 10 datasets sourced from CICIOT2023 containing UDP Flood and Benign DDoS traffic data, and the testing was carried out using three epoch parameters (iterations), namely 10, 50, and 100. The test results using RNN epoch 100 were superior, showing satisfactory performance with an accuracy value of 98%, precision of 99%, recall of 99%, and f1-score of 99%. Based on the experimental results, it can be concluded that combining PCA and RNN is able to detect UDP Flooding attacks by showing high accuracy.