DDOS Attack Detection and Performance Analysis in IOT Network using Machine Learning Approaches

Abstract

The Internet is the most common connecting tool for devices, such as computers, mobile phones, smart watches, etc. These devices communicate with designated servers to provide information. Here we refer to the system that connects numerous autonomous devices known as the Internet of Things (IoT). As the devices are of diverse categories and sometimes very small, it becomes challenging to provide comprehensive security to those in need. However, the sensors on the IoT collect huge amounts of data and the huge network becomes an attractive target for assaulters. One of the several assaults on IoT is Distributed Denial of Service (DDoS). Machine learning can play a crucial role in identifying these attacks in the IoT because of its ability to analyse large amounts of data. Machine learning models can learn the pattern of legitimate traffic and later identify malicious packets that deviate from the learned pattern. Classification techniques can distinguish malicious packets from genuine ones based on several attributes associated with them. This work uses classification techniques such as Random Forest, Gradient Boosting, and XgBoost to determine the malicious packets in traffic. The analysis shows that balancing techniques such as SMOTE and ADASYN are vital in improving the performance of techniques.