Securing the Metaverse: Traffic Application Classification and Anomaly Detection
Vishal Murgai, Venkata Rama Raju Lolabhattu, Roxy Stimpson, Eishita Tripathi, Shiva Chickala · 2024
Metaverse presents a dynamic and expansive digital immersive experience, converges virtual and augmented reality environments to host multiple applications to meet diverse user preferences. As Metaverse continues to grow, robust app detection and network security mechanisms become imperative for safeguarding user privacy, detect cyber threats to bolster trust and confidence. This research proposes a comprehensive app detection and classification mechanism for securing apps within this complex environment. Utilizing a large opensource dataset encompassing diverse network traffic features, we propose a novel approach centered around the classification of Metaverse applications. Our methodology aims to categorize these applications into three distinct groups in Metaverse, namely, network infrastructure, real-time conversational and non-real time apps. Such insights enhances our understanding of their operational dynamics, network planning and facilitating more targeted security measures. Our investigation extends to the realm of anomaly detection, with a particular focus on identifying and scrutinizing aberrant behaviors such as zero-byte packets and irregular traffic patterns. We achieved an accuracy of 85% in app classification with XGBoost and 87% with DNN. Through our analysis, we root-caused “unclassified” app types, attributing their presence to zero-byte packets, indicative of potential port-snooping or Denial of Service (DoS) attacks in the Metaverse realm.