Aprendizado de máquina aplicado à detecção de ataques DDOS com abordagens de inteligência artificial explicável (XAI)

Abstract

The accelerated growth of cyberattacks in recent years has posed a significant threat to the security of computer systems worldwide, especially due to their increasing complexity. Among these threats, Distributed Denial of Service (DDoS) attacks stand out for aiming to make services unavailable by overloading systems with malicious traffic. Given the limitations of traditional signature-based detection methods, this study aims to analyze the application of machine learning techniques for detecting DDoS attacks using Instituto Federal de Pernambuco. Campus Palmares. Curso de Análise e Desenvolvimento de Sistemas. 30 de julho de 2025. a public dataset, as well as to employ explainable artificial intelligence approaches to make the results more transparent and interpretable. The methodology involved the use of the DDoS SDN dataset, with preprocessing steps that included removing null values and irrelevant columns, followed by data normalization using Standard Scaler and One Hot Encoder. The models applied were K-Nearest Neighbors (KNN), Decision Tree, and Random Forest, and were evaluated using metrics such as accuracy, precision, recall, and F1-Score. The results showed high performance from all models, with Decision Tree and Random Forest achieving 100% in all evaluated metrics, while KNN achieved 97.22% accuracy and a 96.43% F1-Score. These findings highlight the effectiveness of machine learning models in identifying malicious patterns and demonstrate their potential to enhance precision, adaptability, and interpretability in the detection of DDoS attacks.