Self-supervised encrypted traffic classification via importance-aware masked clustering

Researchers developed a new method for classifying encrypted network traffic (data traveling across the internet in scrambled form) using self-supervised learning (training an AI without needing humans to label every example). The approach uses importance-aware masked clustering (grouping similar encrypted traffic patterns while focusing on the most informative pieces of data) to identify what type of application or service created the traffic without being able to see the actual content.