Biometric authentication systems are based on comparison of a biometric template, called a reference template, with a fresh one. To increase the security of the authentication phase and the protection of personal data, combining biometrics with a smartcard as another factor of authentication is widely used. A smartcard enables to store the reference template directly in its memory whereas a solution with a central database implies more important constraints on personal data protection. Following this privacy concern, the reference template should not be divulgated outside the smartcard. To this aim, the Match-On-Card (MOC) paradigm has been introduced; in particular for fingerprint recognition. This consists on executing the comparison (also called matching) of two fingerprints directly within the smartcard. Existing MOC solutions have the following problem. Implementation either is software only or uses the arithmetic coprocessor, which is designed to accelerate cryptographic computations but not biometrics ones. In both cases, the trade-off between execution time and biometric error rates is thus not very good. Consequently, adding a security layer, as this is made for cryptographic operations, is hard. In our knowledge, no specific security protections against side channel attacks exist on MOC implementations today. In particular, there exist none security evaluations of the MOC part of a smartcard. Nevertheless, the security of the biometric matching is completely correlated to the privacy of the smartcard owner. In BMOS (Biometric Matching On Smartcard) project, we aim to improve the current situation by designing a hardware implementation enabling a fast and efficient matching, with good biometric performances, while offering the protection of personal biometric data thanks to a matching operation with integrated countermeasures against side channel attacks. BMOS project can be overviewed through three big steps: Analysing and designing a hardware architecture enabling the acceleration of a fingerprint matching with a low level of complexity and great biometric performances; Securing the implementation to achieve a high level of protection of personal data; Validating the effective security by internal evaluation. With their respective expertise in, biometrics, smartcard software security, embedded software security, electronic circuits countermeasures, the partners will contribute to handle the following technical and scientific problems: vulnerabilities evaluation of MOC against side channel and fault injection attacks; scientific research of needed countermeasures ; design and implementation of an architecture with low complexity fitting to smartcard constraints while offering high biometric performances and low execution time. Security of hardware implementations is today mainly focused on cryptographic operations and only a few on biometric ones. Therefore, beneficial effects of BMOS project will be important on a scientific side as on an economic side. BMOS project wants to create a technological breakthrough which could be exploited in a short term: a MOC technology, which would be efficient and ensuring protection of personal data, represents an important factor of growth of biometric authentication solutions, particularly for ID documents market.

Nowadays embedded systems are pervasive, as they can be found in banks, health, biometric passports or smartphones. Therefore ensuring the security of those systems and their applications constitutes a critical requirement for maintaining functionality and data protection. Specific designs and countermeasures are integrated by developers and manufacturers to enhance the resistance of their product against attacks. The security level of the products are then tested and evaluated by experts. Regarding vulnerability analysis, those evaluations must match the current state of the art. The fault injection related state of the art require maximizing the exhaustiveness of the attacks analysis and the forthcoming multiple attacks---both spatial and temporal---will further complicate the analysis. The objective of the SERTIF project is to rationalize and automate as much as possible the robustness assessment process of highly secure systems against fault injection, starting from the code analysis and ending with the physical realization of attacks, with the aim of taking into account multiple faults, which is presently a significant lock. To achieve this goal the SERTIF project will respond to the following challenges: 1) characterization of fault models truthful to physical attacks and implementation of those models in code analysis ; 2) definition of coverage criteria to ensure the completeness of a fault injection campaign with respect to security targets ; 3) aid to the development of secure application by analyzing the effectiveness and relevance of the countermeasures in code ; 4) transition to multiple faults attacks. The techniques used will be low-level simulation, code mutation and more generally the combination of static and dynamic analysis on low-level code to handle the combinatorial explosion and to qualify the results of the vulnerability analysis against fault injection. This project is based on the great complementary of the partners, who represent the various actors involved in the certification and development process of embedded applications on secure components. Furthermore the partners are already independently developing their own tools for vulnerability analysis. The Vérimag laboratory is the project leader and a well-known academic actor in the domain of embedded systems verification and validation. Vérimag developed strong skills in the security field (protocol analysis, vulnerability detection in low-level code, security and certification of smart card applications). The CESTI-LETI is an electronic component and embedded software ITSEF recognized by ANSSI in the context of the French Evaluation and Certification Scheme. The CESTI-LETI is a specialist for the most critical objects of the state of the art (smart cards, integrated circuits, banking applications, biometric passports, ...). Morpho is a pioneer and market leader in the field of e-Documents like smartcards. Morpho develops secure products for telecommunications, health, identification, transport, pay TV and payment sectors. The Morpho's team working on this project is in charge of smartcard security. Coordinators : Project leader and coordinator for Vérimag: Marie-Laure Potet Coordinator for CEA-LETI: Jessy Clédière Coordinator for Morpho: Than-Ha Le The project should result in a significant skills improvement for the involved partners and the research field, regarding techniques and methodology for the evaluation of the robustness of embedded applications against fault injection. Moreover, the SERTIF project aims to build a public benchmark of applications relevant to the field, specifically protected against fault injection. This benchmark will allow for public comparison of the analysis results and could constitute a step forward for the entire research field.

With the exponential growth in the use of digital identity over the Internet and across country borders, the abundance of identification numbers and login/passwords induced by modern societies, biometric identification, combining practicality and simplicity for users, is booming. This is particularly noticeable on mobile devices. In the meantime, fraud keeps developing and countermeasures to spoofing are unsatisfactory. That’s why innovative security solutions are needed to protect citizens from this specific threat on their identity. In this project, we focus on 3D face modality and propose to study and develop an active sensor for biometric acquisition on smartphones based on diffractive optics technology. The miniaturized optical system to be designed in LIVEFACE will enable 2D/3D face image acquisition and liveness detection on mobile devices. We will develop algorithms dedicated to 3D face acquisition and reconstruction in 2D/3D for mobile devices, taking into account the baseline between the camera and the projector as well as the projected patterns. We will then take advantage of multi-facial information in order to propose more accurate and more secure recognition algorithms, thanks to the use of multi-modalities. We will study face tissues properties and calculating new specific characteristics for liveness detection; then design an efficient anti-spoofing algorithm using unique features and different fraud fighting methods. The study of optical skin properties will exploit the patterns projected on the face together with an appropriate choice of wavelength. Tests and evaluation of the prototype will be done on a representative number of lures and real faces. Finally, an analysis of the legal frameworks applicable to anti-spoofing data and to the technology developed will be conducted.