New research on biometric reading of mobile fingerprints, synthetic data presented at eu-LISA summit

0

Among a series of presentations on technologies that are or could be used at the European border during a eu-LISA Industry Roundtable held last week in France, two biometrics research projects, one involving hardware and the other involving software, stood out.

Javier Galbally from eu-LISA chaired the industry presentations session on “Biometric Solutions and Business Processes for Passenger Processing at Land/Sea BCPs”, which started with two presenters from the French Judicial Police and Isorg sharing their work on mobile biometric capture.

French police want mobile fingerprinting capability in part to collect biometric data from crime victims, to differentiate it from other data, such as latent fingerprints found at a crime scene. The identification of corpses, verification of identity and right of residence and registration in information systems for minor offenses also motivated the project.

Jennifer Aflalo, Isorg’s Biometric and Identity Security Program Manager, explained Isorg’s organic photodiode sensors, along with sensor implementation and image quality considerations. Used in a scanner that provides a dedicated light source, Isorg’s technology meets high-security fingerprint quality criteria, says Aflalo.

The initial prototype design was followed by a larger surface version and feedback from French law enforcement authorities, then a smartphone-sized system designed for final evaluation and comparison with other other market alternatives.

Good quality images were delivered with minimal processing, marking a success for Isorg, according to the presentation, with results similar to other FBI-certified biometric technologies.

As an optical device, the scanner can also capture ID documents and even latent prints, says Aflalo. Tests showed that latent prints could be collected on the device, then prints of living subjects, without the quality of the images of the latter being affected for up to 100 repetitions.

A major benefit of the test from Isorg’s perspective was the work of setting up the image processing needed to meet police requirements.

Synthetic data

Marcel Grimmer, researcher at NTNU, presented a project carried out with eu-LISA to evaluate and improve the synthetic biometric datasets used by the regional agency.

The synthetic biometric datasets were composed to provide the necessary volume of training data without encountering data access and privacy issues.

The test aimed to determine whether the synthetic faces exhibit similar characteristics to authentic samples, to assess their quality and to compare the scores obtained with open source facial recognition systems. To do this, the researchers evaluated images of unmated synthetic faces, modified facial attributes to generate mated synthetic samples, and compared the mated faces to real data.

Generative adversarial networks (GANs) have been used for modifications such as changing the apparent age of the face.

Grimmer reviewed the rapid advances in the capabilities of these GANs, drawing the conclusion that the synthetic subsets each have similar biometric quality to real-life reference data.

The mating comparison had a much lower degree of similarity than the mating comparisons, especially at extreme ages.

Grimmer concludes that face-aging algorithms appear to be effective in generating training data, in terms of biometric quality. Further experiments on robustness to larger age differences are recommended.

Some challenges need to be resolved before synthetic data is suitable for use as the sole tool for evaluating operational systems, despite the encouraging results, Grimmer says.

Article topics

biometric research | border control | Entry/Exit System (EES) | eu-LISA | European | facial biometrics | fingerprint biometrics | isorg | mobile biometrics | synthetic data

Share.

About Author

Comments are closed.