Remote mobile biometric authentication system for unspecialised mobile devices (BioMobi)

Demand for remote mobile authentication systems continues to grow. The most popular method of authentication using a PIN has multiple drawbacks and has recently been widely supplemented by biometric chips embedded in mobile devices.

In the project, our goal was to build such a system that would apply to commonly available devices, without the need for special biometric sensors. It was important to us that the solutions be adaptable to security needs and user preferences through the use of a multibiometric system, i.e. having the ability to measure multiple biometric modalities.

The BioMobi system we developed is thus a response to the shortcomings of existing mobile authentication programs that use biometric techniques in the identity verification process.

The BioMobi project concluded with the development of a system that uses biometric identity identification methods and accounts for the specificity of biometric samples obtained from mobile devices.

We selected the biometric techniques that we applied to the mobile application. They use both classical and original methods of deep neural networks and artificial intelligence. We researched in-house solutions and improved existing ones. We devoted a lot of attention to the search and construction of databases necessary for proper learning of deep neural networks. We classified the obtained solutions in terms of recognition quality, execution time, required computing resources, convenience of use, etc. For each biometric modality, we developed at least one solution that received a positive rating on biometric data obtained from external sources.

We also paid a lot of attention in the project to preventing biometric sample presentation (PAD) attacks. We considered several proposals for preventing attacks. We tested the algorithms under different conditions and identified the most effective ones.

In the course of our work, we took into account the legal environment paying special attention to the Regulation on the Protection of Personal Data (RODO/GDPR).

The results of the project can be useful in building biometric authentication systems resistant to attacks.

Jarosław Wójtowicz, the project leader, says this about the project:

The idea behind the project was to create a multimodal biometric security system independent of manufacturers of specific biometric sensors and closed proprietary solutions. By design, such a system is easier to be used widely, while meeting manufacturer-independent security requirements.

The project was funded under the CYBERSECIDENT/382354/II/NCBR/2018 competition and ended in September 2022.

Our partners were the AGH University of Kraków and the Łódź University of Technology.