Anticipating DDoS Incursions in Software-Defined Networking Using Explainable AI and Federated Learning

Abstract

Distributed Denial-of-Service (DDoS) infiltrations being an escalating menace to Software-Defined Networking (SDN) infrastructure. It is vital to have effective mechanisms in place to predict and counter these cyberattacks. In this thesis, we recommend an innovative methodology to anticipate DDoS incursions in SDN using Explainable AI (XAI) and Federated Learning (FL). Our novel framework employs XAI techniques to enable network administrators to decipher the results of the prediction with precision. FL is integrated to train machine learning models using data from diverse sources, while prioritizing data privacy to safeguard confidential network information. Additionally, we explore various XAI techniques and analyze their effectiveness in interpreting the results of our models. Our proposed framework was evaluated using real-world network traffic datasets, and its performance was compared with existing techniques. Results revealed that the amalgamation of XAI and FL techniques surpassed current methods in terms of predictability and interpretability. We demonstrated the effectiveness of different XAI techniques in producing explicable results that can assist network administrators in identifying the causes of an attack and formulating effective countermeasures. Overall, our innovative framework provides a promising solution for anticipating DDoS incursions in SDN infrastructure. It serves as a valuable tool for network administrators in detecting and neutralizing cyberattacks. The framework employs XAI and FL techniques to achieve precise, efficient, and explicable DDoS incursion anticipation in SDN. This thesis contributes to the field of SDN security by proposing a solution that can safeguard SDN networks from DDoS incursions.