Biometrics in mobile environments
The amount of sensitive data that needs to be protected, not only at institutions or companies levels but also for ordinary people, is increasing exponentially [10] [23]. Nowadays, it is common to use the smartphone to access bank accounts [20], make payments or handle important information in general [7, 8], which leads to the necessity of increase the security in those devices [17, 24]. Usually, the applied methods to assure security in mobile devices are based on PINs or passwords, which can be easy to forget and forge, so that, other approaches are arising. In particular, biometric recognition is suggested to be embedded in mobile devices for many reasons. The first one is the large amount of devices already deployed, which has reached the situation that it is difficult to find someone that does not possess and use daily devices such as smartphones or tablets. The second one is that for some biometric modalities, the capture device is already included within the mobile device (e.g., camera for face recognition, touch screen for handwritten signature recognition, microphone for speaker recognition, or the inclusion of some swipe sensors for fingerprint verification). Handwritten signature, voice and face recognition has been suggested as the most suitable modalities [19].
This leads to an important reduction in the cost of the deployment, as users already have those devices and they should only acquire the application. Other important factors are the necessity of having ID portable devices by security forces (e.g., for suspects identification) or for signing documents on the spot. Also, as users are already familiar with this kind of devices, the usability level achieved could be improved, although, as it will be mentioned below, mobility also creates new usability challenges. Due to marketing needs, mobile devices are improving every day, which will allow powerful biometric algorithms in the near future. As an important drawback, mobile devices present security concerns related to how the operating system controls the way that installed applications access memory data and communication buffers. A lack of a strict control compromises the integration of biometrics as sensitive data may be endangered.
Usability and accessibility concerns
One of the major drawbacks when using biometrics in mobile devices is the lack of usability being this technology a challenge for users in many cases. Almost all the work done in biometrics is devoted to improve algorithms performance and bringing the Equal Error Rate (EER) close to zero. But while this kind of research is necessary, working on improving user interaction with systems is also extremely important, as a lack of usability could mean not only the rejection of the system by the users, but also a reduction in the expected performance of the biometric system. In order to increase the easiness and encourage the use of biometrics it is necessary to improve its usability and accessibility, making it reachable for a wider percent of population.
One of the collectives usually excluded at the time of design security systems is the disabled people, who are around 15 % of the world population [35]. Furthermore, it is important to highlight that every individual is potentially dependent (illnesses, age, pregnancy, etc.). Improving biometrics designs would be beneficial not only for disabled people but for many others who find the technology complicated to use. It could be thought that biometric recognition is challenging for disabled people but we show in this work that a correct design can make the process easy for everyone. Specifically, we focus on face recognition for visually impaired users, providing audio feedback and instructions. Face recognition has shown to have a good acceptation and it is one of the less intrusive modalities for users. There are several works in the literature embedding face recognition in mobile environments [4, 29, 30]. Nevertheless, the amount of works in biometrics accessibility is scarce yet. One of the first approaches is a universal access control to mobile devices through fingerprint and handwritten signature developed by authors [22]. Other recent works in biometrics accessibility show the advantages against other alternatives such as PIN or passwords [37]. These researches point the necessity of reliable solutions which could ease several procedures to people with accessibility concerns.
State of the art in biometrics usability
There are several usability works in biometrics in the literature and most of them come from the usability definition given by the ISO 9241:2010 [1]: “The extent to which a product can be used by specified users to achieve specified goals with effectiveness, efficiency and satisfaction in a specified context of use”. The National Institute of Standards and Technology (NIST) made some experiments emphasizing in ergonomics to better capture user traits. For instance, in [27] an experiment analysing the optimal device position regarding height or in [28] where they measured the usability of the face image capturing system at the US ports of entry. One of the first usability studies in biometrics was an enrolment trial in the UK [3] conducted by Atos. Kukula et al. designed a model, the HBSI (Human Biometric System Interaction) [15], where the interaction between the user and the system is studied through ergonomics, usability and signal processing. In Kukula et al. [33] analysed the different kinds of possible errors when applying hand geometry recognition. Another example is [16] where an extensive analysis on the fingerprint devices ergonomics is done.
Authors have carried out several experiments analysing various factors which affect usability in biometrics. In [5] the use of different styluses in dynamic signature verification on an iPad was studied reaching interesting outcomes regarding ergonomics. The stress as a key usability factor in biometric recognition was analysed in [6], where authors showed that mobile biometrics are reliable even when the user is under stressful situations (banks, shops, post office, etc.). The most relevant works providing feedback to users in biometrics are the experiments made by NIST. In [11] and [9] they present a quality-driven interactive real-time user feedback mechanism for unattended fingerprint kiosk, where the application shows pictures (visual feedback) to users helping them to better place their fingertips on the sensor. In our case, we suggest a new feedback mode providing audio feedback in order to guide visually impaired users through the biometric process.
Study of the time evolution
Several visually impaired people participated in this usability evaluation of face recognition, where they were asked to take self-photos with a mobile device. We prepared different experiments with different kinds of feedback divided into two sessions. Once all the images had been taken we analysed the face recognition performance in contrast with the time employed in the process. In mobile environments, one of the most critical aspects is the time employed in the authentication process because long times would lead to users rejection and/or security concerns. On the contrary, quick interactions could involve misuses and errors in recognition. Then, in this work we have focused on the efficiency (as defined by ISO 9241:2010, the time spent in tasks) of biometrics in mobile devices.
There are not too many works on accessibility in biometrics [21] and there is not a standard methodology yet. In this work, we compare instructions with audio feedback in real time following the state of the art in interfaces [31] accessibility and applying them to biometrics [2]. This work comprises an extensive analysis of the time influence in face recognition for visually impaired users both in performance and usability, obtaining several important outcomes regarding: the importance of the received feedback, the variability of the performance and the usable images rates along with the time spent in the process. This paper is divided as follows: in “Evaluation set up” the evaluation set up is provided. We explain the methodology and the experimentation in the “Methodology and experimentation”. The results are in “Results and discussion” and the conclusions and future work are in “Conclusions and future work”.