Automatic facial recognition system for continuous authentication of users in service
The aim of the SMOWL project is the generation of continuous user identity authentication software during online training activities, detecting any type of attempt at identity theft or incorrect behaviour.
Vicomtech’s role is that of implementing identification software based on the extraction of the identification vectors of the users and the comparison with said information previously stored in the database.
The implementation of a facial DETECTOR in uncontrolled environments which will use Machine Learning paradigms capable of solving detection in variable lighting conditions, different poses, accessories such as glasses/caps or partial obstructions would be a challenge for technology. Likewise, the implementation of a facial RECOGNISER in uncontrolled environments which will use Machine Learning paradigms capable of eliminating gestures, expressions, undesired effects due to accessories and possible changes of appearance which make identification against a database difficult. The recogniser must verify the detected face against a group of images, to see whether that face belongs to the expected person.
As a result of the project, software has been obtained which allows user identification, with high rates of success in the recognition of imposters and with acceptable response times for the existing scales. Currently, Vicomtech is still working to improve the results of said software.