Biometric-based identity management is becoming more and more popular. Alongside classical applications like biometric passports, border monitoring and access control, biometric data are increasingly used to access cloud-based services, wherein users' recognition needs to be carried out by mobile devices.
Though desirable for the many benefits it may bring, the diffusion of biometric based recognition is proceeding at a much slower pace than expected. In fact, despite most research has focused on performance analysis and investigation of methods of biometric acquisition and feature extraction, only little attention has been paid to security and privacy threats.
A serious and very general concern regards the possibility that a biometric template is stolen, e.g., because the database with the biometric templates of enrolled users is violated. Due to the non-revocability of biometric templates, theft of biometric data may have serious consequences, which cannot be easily remedied. Moreover, the use of biometric data poses serious privacy concerns, given the indissoluble link with the data owner. Unluckily, privacy protection is rarely addressed in most practical systems, even if addressed from a theoretical point of view in several studies. As a matter of fact, concerns regarding the privacy of users are among the most important factors precluding the acceptance of biometric systems in our society.
Due to their importance, several solutions have been proposed to cope with the above problems; however, no fully satisfactory solution exists especially when the biometric recognition system operates in a mobile environment to ensure access to cloud computing and storage services. In addition, the great majority of research has considered each of the above threats by themselves providing stand-alone solutions, whose integration into an overall system taking into account all possible security threats is not easy, if at all possible.
SPoB-MA aims to provide innovative secure multi-biometric protocols, optimized to allow efficient user authentication through mobile phones. Results obtained with single biometrics in literature pave the way for secure multi-biometric protocols. Selective biometric protection is instead an innovative idea where some private information is disclosed to provide higher efficiency. Despite the obvious feasibility, in-depth studies are necessary to measure and quantify the privacy loss introduced.
WP0. State of the art analysis. A complete analysis of the state of the art in MPC will be performed to identify the most suitable cryptographic primitives and existing protocols for secure biometric identification. Moreover a study on the recent trends in (non-secure) multi-biometric authentication will be performed. The WP aims to provide the needed knowledge to all SPoB-MA participants and better define other WPs' activities.
WP1. Multibiometric authentication in the encrypted domain. The most suitable cryptographic protocols for privacy preserving biometric authentication are selected. The choice will be first of all between interactive and noninteractive primitives. While the latter is really appealing for mobile devices, thanks to the possibility of moving all the computation on the server side, its complexity is higher. WP1 will consider different security models and the possibility to involve non-trusted third parties in the computation [4]. This complicates the protocol because parties can collude, but also allows to move expensive computation outside mobile devices. WP1 also aims to provide protocols for secure multi-biometric matching characterized by a good trade-off between efficiency and accuracy. Optimization can be performed at different levels, the most explored one being the optimization of the MPC primitives. A less explored, but equally viable, strategy relies on the optimization of the aspects of the algorithms to be implemented in a SPED fashion. Being accuracy of single biometric protocols often considered to be sufficient in mobile devices, our multi-biometric protocol does not aims to provide higher accuracy, such as in [3], but same accuracy of single biometric protocols with lower complexity.
WP2. Selective biometric protection. WP2 will explore the idea of increasing efficiency by protecting different parts of the biometric template with solutions having different security levels. During enrollment, after having acquired the biometric, features are extracted and split in two parts: one protected with strong cryptographic tools, while the other one protected by using lighter cryptographic mechanisms, or left unprotected. Being different secure protocols involved, the strategy for data splitting and assignment to different protection techniques is not obvious to determine. In same cases it is easy to identify the part of the information that is more important to be protected (i.e. in iris identification protocol making use of masks to localize and discard from the matching protocol the sections of the iris affected by noise, the iris template must be protected with heavy cryptographic tools, while the mask could be protected with lighter mechanisms), in other cases, if all the features have the same importance and it is not simple to define the part of them that could be leaked, feature vector could be split according to template geometry, or by using stochastic algorithms.
WP3. Realization of multi-biometric database. While several biometric databases are available, few databases contain more biometrics provided from the same people. For this reason WP4 aims to build a database containing face, iris, fingerprints and other biometrics that can be collected by using standard mobile devices. The database will be used for validating the results obtained by previous WPs and will be made available to the scientific community.
Expected impact of SPoB-MA
The results obtained by the project will contribute to advance the state of the art in privacy preserving biometrics authentication, thanks to the innovative ideas and the PI expertise in the SPED field and its application to biometric protocols.
SPoB-MA, with the development of SPED-based solutions is expected to close the gap between theoretical studies on MPC and homomorphic encryption and practical systems. Despite SPED technology has long been been advocated as a perfect solution to protect biometric data, in fact, no practical system based on such a technology has been developed yet.
Above all, selective biometric protection is a widely unexplored field that has the potentiality to revolutionize biometric security research.
SPoB-MA can be of interest to many potential users, in multimedia management and security in general and biometrics in particular. Many companies can benefit from remote authentication protocols able to recognize their customers, independently from the device used, while protecting user privacy, in respect of GDPR.