QUICPro: Integrating Deep Reinforcement Learning to Defend against QUIC Handshake Flooding Attacks
Y. A. Joarder, Carol Fung · 2024
In recent years, QUIC protocol has emerged as a promising alternative to traditional transport protocols like TCP and UDP, offering significant performance improvements in latency and throughput. Additionally, QUIC provides strong security protection mechanisms. However, like most other new protocols, QUIC also faces new security challenges. In our previous study, we found that handshake flooding attacks can exploit vulnerabilities in the QUIC handshake process to overwhelm server resources and disrupt service availability. In this lightening paper, we present QUICPro, a novel approach that leverages Deep Reinforcement Learning (DRL) techniques for dynamic network security optimization to enhance QUIC protocol security against handshake flooding attacks. By integrating DRL algorithms with adaptive defence mechanisms, QUICPro offers a proactive and adaptive solution capable of detecting and mitigating handshake flooding attacks in real time. This paper analyzes the QUICPro framework, highlighting its technical components, implementation details, and expected outcomes.