A Model for the Detection of Denial-of-Service Attack Patterns

Abstract

Denial of Service (DoS) attacks is a critical cybersecurity threat that can disrupt network services by overwhelming it with illegitimate traffic. This paper presents a hybrid detection and mitigation system to address DoS attacks. The system leverages Python as the primary programming language, incorporating its robust ecosystem of libraries such as Scikit-learn, TensorFlow, and Tkinter for machine learning, feature extraction, and GUI development. The system integrates a hybrid model combining K-means clustering with a Random Forest classifier. Initially, the Kmeans algorithm groups data into clusters, which are then fed into the Random Forest for classification. The system was trained and evaluated using an undersampled DoS dataset to ensure balanced class representation. Results show that the model performed quite excellently, achieving a 98.97% accuracy in detecting DoS attacks, compared to other models such as XGBoost (98.48%) and multidimensional feature approaches (98.96%). The model’s deployment on a web-based platform demonstrated its ability to filter and classify incoming network traffic into normal or attack types like SIDDOS and UDP-Flood in real-time, effectively mitigating the threats posed by DoS attacks.