DoS Attack Detection using Machine Learning by Ahmed Badeeah and Rafael Uribe

The primary objective of this project is to develop a machine learning model capable of classifying network traffic data into two categories: good traffic and bad traffic. The model will be trained using a network traffic dataset, which contains traffic-related features such as packet size, traffic duration, etc. This project's importance lies in the increasing need for network security with the proliferation of internet-enabled devices and their interconnectivity. Attackers have become more sophisticated in their attempts to obtain sensitive information or compromise critical systems. Our project aims to investigate the use of machine learning techniques to classify network traffic data, providing network administrators with an efficient means of detecting active Denial-of-Service (DoS) attacks.

Our project has several sub-problems that need to be addressed, such as data preprocessing, exploratory data analysis, feature selection, and model selection. We have reviewed several studies conducted on the classification of network traffic data using machine learning techniques. We have identified some popular machine learning models for this task, including logistic regression, decision trees, random forests, and support vector machines. After analyzing the data, we will select the most relevant features for classification, evaluate different machine learning models, and determine the most appropriate model for this classification task.

The project's system architecture includes an attacker's computer, a victim's computer, a detection computer, and a router. To implement aggressive DoS detection, we use PyShark, a Python wrapper around the Wireshark network capture tool. PyShark allows us to capture network traffic in real-time and flag it as benign or malicious. We use machine learning models to analyze reported traffic and detect active DoS attacks. Our experiments show that the system can detect active DoS attacks with a high accuracy rate of over 95%.

We are confident that this project will provide a comprehensive guide to building a proactive DoS detection system using machine learning techniques. This project's implementation details can dramatically improve network security by accurately identifying and mitigating malicious traffic in real-time.

Thank you for your interest in this project.

Sincerely,

Ahmed Badeeah and Rafael Uribe