Refining Intrusion Detection in the Age of QUIC: Leveraging Machine Learning for Enhanced Security

The emergence of the QUIC protocol could change online traffic patterns significantly. It encrypts transport layer data, potentially affecting how we analyze internet traffic and detect intrusions. We propose using supervised machine learning to spot bot-generated flows during DDoS attacks, focusing on the challenges posed by QUIC. Machine learning has proven its mettle in numerous research endeavors aimed at pinpointing network intrusions, yielding promising results. Nonetheless, as we delve deeper into this realm, it becomes imperative to consider the evolving landscape of network traffic. With the advent of technologies like the QUIC protocol, which encrypts transport layer data, the traditional methods of intrusion detection may face challenges. Thus, while leveraging machine learning remains a potent strategy, it’s essential to adapt our methodologies to accommodate these novel developments. Our work involves a multifaceted approach comprising several stages: initially, simulating real-world traffic scenarios to capture diverse data patterns; followed by the creation of labeled datasets, enabling a comprehensive understanding of data quality and characteristics; and finally, conducting experimentation with a range of machine learning algorithms to discern the most suitable for predictive modeling.