%0 Journal Article 
%A Monschein, David
%A Knecht, Markus
%A Al Samara, Mustafa
%A Peregrina, José Antonio
%A Piotrowski, Tim
%A Gohlke, Sascha
%A Bennis, Ismail
%A Azan, Wilfrid
%A Abouaissa, Abdelhafid
%A García de Prado Fontela, Alfonso
%A Gilg, Marc
%A Neumann, Eva Maria
%A Nochta, Zoltán
%A Ortiz Bellot, Guadalupe
%A Szilagyi, Ioan
%A Waldhorst, Oliver P.
%A Zirpins, Christian
%T A Privacy-Preserving Behavioral Authentication Framework for Public Transport Applications
%D 2026
%@ 2169-3536
%U http://hdl.handle.net/10498/39748
%X Public transportation systems increasingly rely on mobile services for ticketing, trip planning, and account management. These systems require secure and user-friendly authentication because traditional login mechanisms often disrupt the travel experience and only protect the initial login. In this context, behavioral authentication based on interaction patterns and sensor data offers a promising solution for continuous user verification. However, its practical deployment raises challenges related to privacy, robustness, distributed model training, and integration with existing authentication infrastructures. This paper presents the design and deployment of a privacy-preserving behavioral authentication framework for public transportation applications. The framework focuses on system-level integration of established privacy-preserving machine learning and authentication techniques. It combines client-side preprocessing and homomorphic encryption of behavioral data with server-side machine learning-based analysis. It further incorporates model optimization techniques to improve robustness against impersonation and supports collaborative model training across providers through federated learning with coordinated preprocessing and data governance. To enable practical deployment, the framework integrates behavioral authentication into existing OAuth 2.0 and OpenID Connect-based infrastructures, including authentication flows, token management, and cross-provider interoperability.We evaluated the proposed framework through preliminary experiments on a public dataset, a field test with a real-world public transport application, and a second field test exploring federated learning in a collaborative training scenario. The results demonstrate the practical feasibility of the proposed architecture and provide insights into the challenges of deploying behavioral authentication in real-world mobility ecosystems.

%K Behavioral Authentication
%K Behavioral Biometrics
%K Privacy-Preserving Machine Learning
%K Real-World Deployment
%K Federated Learning
%K Homomorphic Encryption
%K Authentication Infrastructure
%~ Universidad de Cádiz