Privacy Attacks and Defenses for Digital Twin Migrations in Vehicular Metaverses

TL;DR

This paper investigates privacy risks during digital twin migrations in vehicular metaverses and proposes dual pseudonym schemes and synchronization frameworks to enhance privacy, demonstrating significant utility improvements over existing methods.

Contribution

It introduces specific privacy attack models for VT migrations and proposes novel dual pseudonym and synchronization schemes to defend against these attacks.

Findings

Privacy attacks during VT migrations identified and modeled.

Proposed pseudonym schemes significantly improve privacy entropy.

Utility of VMUs increased by 33.8% with the new schemes.

Abstract

The gradual fusion of intelligent transportation systems with metaverse technologies is giving rise to vehicular metaverses, which blend virtual spaces with physical space. As indispensable components for vehicular metaverses, Vehicular Twins (VTs) are digital replicas of Vehicular Metaverse Users (VMUs) and facilitate customized metaverse services to VMUs. VTs are established and maintained in RoadSide Units (RSUs) with sufficient computing and storage resources. Due to the limited communication coverage of RSUs and the high mobility of VMUs, VTs need to be migrated among RSUs to ensure real-time and seamless services for VMUs. However, during VT migrations, physical-virtual synchronization and massive communications among VTs may cause identity and location privacy disclosures of VMUs and VTs. In this article, we study privacy issues and the corresponding defenses for VT migrations in vehicular metaverses. We first present four kinds of specific privacy attacks during VT migrations. Then, we propose a VMU-VT dual pseudonym scheme and a synchronous pseudonym change framework to defend against these attacks. Additionally, we evaluate average privacy entropy for pseudonym changes and optimize the number of pseudonym distribution based on inventory theory. Numerical results show that the average utility of VMUs under our proposed schemes is 33.8% higher than that under the equal distribution scheme, demonstrating the superiority of our schemes.