Human beings have many more difficulties in identifying false faces than computers that make facial recognition, with less accurate this task.
Así se desprende de un nuevo estudio publicado por ID R&D, proveedor de biometría facial y de voz basada en la Inteligencia Artificial (IA) propiedad de Mitek Systems, titulado Humano vs Máquina: ¿Pueden las personas detectar las falsificaciones mejor que la Inteligencia Artificial?, en el que compara la habilidad de personas para detectar la vitalidad facial, es decir, identificar si un rostro es o no real.
The study has tested humans and machines presenting the most used impersonation techniques: printed photographs, videos, digital images or using a 2D or 3D mask.
Humans competing with computers
Computers have surpassed humans in the five techniques evaluated by obtaining an error percentage of 0% in 175.000 images and in all types of attack.
People, on the contrary, have obtained a much lower degree of precision for each type of supplant technique, including the erroneous identification of 30% of the falsificifications of printed photos, one of the easiest type of attack to perform by the scammers.
Even when a group of seventeen people deliberated on the images, giving a more precise result than an individual human, their majority decisions were never better than the performance of the computer in the same task.
People are slower than machines
Humans took an average of 4.8 seconds per image to identify life tests, while computers that work with a CPU nucleus recorded an average of less than 0.5 seconds per image.
In this report, IA has erroneously classified only 1% of authentic faces as false.On the other hand, humans erroneously classified 18% of authentic faces as false.