ETD

• authentication
• cyber-physical system
• formal verification
• intelligent transportation system
• machine learning
• Proverif
• security

Identity is not merely a label, such as a name or physical attribute, but a set of interrelated features and relationships that evolve with purpose, behavior, context, environment, and time. In the digital domain, this understanding underpins authentication, the process of verifying identity, which can evolve beyond static credentials to incorporate contextual, temporal, and physical-layer evidence.
Cyber-Physical Systems (CPSs) exemplify this challenge, as their identities emerge from the interaction between cyber components and physical processes. For example, Intelligent Transportation Systems (ITSs) provide a critical case study, where road vehicles operate as CPSs in dynamic, safety-critical environments, making authentication, and consequently trust, critical. This work suggests that authentication in road vehicles can be strengthened by combining digital credentials with physical factors and evidence drawn from the real-world context of interactions, e.g., by linking identity to both location and timing. This work proposes a context-based Multi-Factor Authentication (MFA) protocol that integrates cryptographic assurances with physical evidence across heterogeneous communication channels. The protocol is modeled and formally verified with ProVerif, demonstrating resilience against Dolev-Yao and distance-bounding adversaries. To assess real-world feasibility, a prototype is implemented in which vehicle physical elements, such as headlights, transmit optical challenge–response sequences that are decoded by onboard cameras and classified using Convolutional Neural Networks (CNNs). Real-car testbeds validate robustness under varied conditions, achieving high accuracy and low latency. The findings show that contextual evidence combined with cryptographic techniques can improve resilience to remote and proximity-based threats. More broadly, this work aims to explore how authentication in CPS can be understood and implemented as a dynamic, context-aware process, offering a preliminary step toward more secure and trustworthy infrastructures where non-human agents operate interdependently.