Periodic Reporting for period 2 - AMBER (AMBER - enhAnced Mobile BiomEtRics)
Periodo di rendicontazione: 2019-01-01 al 2020-12-31
As a Marie-Curie ITN, AMBER’s objectives are as follows: (1) address a range of current issues facing biometric solutions on mobile devices requiring timely research and development; (2) collate Europe-wide complementary expertise to investigate these issues and provide a structure and environment to effectively facilitate training; (3) train and equip the next generation of researchers to define, investigate and implement solutions , and provide transferrable skills to enable effective planning, management and communication of research ideas and outcome.; (4) develop solutions and theory to ensure secure, ubiquitous and efficient authentication whilst protecting privacy of citizens.
The first technical work package looked at usability performance and reliability authentication of biometrics on mobile platforms, investigating how different real-life scenarios such as using a device whilst walking or on public transport affect user identification, distinguishing human from fake interactions, and the effect of an individual’s circumstances on identification (e.g. effect of diabetes on iris recognition). The work led to the development of novel authentication solutions that are applicable to a range of industrial settings such as banking, and contribute towards improving current evaluation methodologies and risk analysis standards and help define new guidelines in making biometrics more accessible to the population.
The second technical work package assessed implementations for exploiting continuous authentication on mobile devices and multimodal architectures and producing, amongst other outputs a key piece of work on finger vein identity as a biometric authentication system that can be applied to mobile devices and has both industrial and economic applications.
The third technical work package focused on privacy protection issues of biometrics on mobile platforms. The research resulted in a number of recommendations such as the exploitation of homomorphic encryption techniques to enhance security, development of presentation attack detection methods, suggestions for a comprehensive privacy requirements model and improvement of privacy in the case of voice activated home assistants.
1. Enhancing user identification and authentication on mobile devices under various conditions including different environments, presentation attacks, and physical characteristics of the users themselves. The results went beyond the state-of-the-art in that: (i) a multi-modal solution was developed that enhances security of identification compared to current methods and is applicable to industrial settings such as banking; (ii) investigations into factors affecting recognition by iris, face and voice identification methods confirmed that health (e.g. diabetes for iris recognition), (iii) use of make-up and differences in mood and time of day for voice recognition significantly affected identification; (iv) a robust model for expected performance of mobile biometric systems across different usage environments was established; (v) algorithms that improved mobile face and speaker recognition were developed increasing the robustness of the system’s security.
2. Solutions for mobile biometric interactions investigated how biometric data is collected and modelled. AMBER also studied biometric identification and biometric verification using finger vein templates. This resulted in a key piece of work recommending that finger vein identification systems should be based on sequential image templates, using a mobile platform as a temporary paired solution.
3. Privacy, security and confidence in mobile biometric interactions were investigated by examining methods and policies at both theoretical and practical levels and resulted in a number of developments: (i) exploiting homomorphic encryption techniques with regards to template protection on mobile devices was illustrated as one of the best solutions to enhance security as multi biometrics and cloud computing gain momentum in the mobile environment, (ii) presentation attack detection methods were developed contributing to the security of biometric systems, (iii) suggestions for a comprehensive privacy requirements model were devised and tested, and (iv) work was carried out to improve privacy in the case of voice activated mobile devices (e.g. intelligent voice home assistants) to ensure they are designed on the principles of privacy and security.
The work performance by AMBER ESRs has significantly contributed to solving challenges in the use of mobile biometric systems. The Importance and relevance of the work was assured through the involvement of industrial partners who are also best placed to apply the solutions and recommendations developed. Importantly, improvements in biometric security and privacy will increase users’ confidence in their mobile and voice activated devices.
AMBER also had a significant impact on the employability of the ESRs working within the network by providing them with a dedicated training programme including secondment to a second academic institution and an industrial placement.