If you’ve watched the movie Catch Me If You Can, then you might wonder how the main character Frank Abagnale Jr (played by Leonardo DiCaprio) could get away with impersonating a pilot and stealing over US$2.8mil in payroll cheques from the company.
Besides such scams, there are other frauds that happen in the travel industry, such as passport forgeries. Face morphing is a method used by fraudsters where photographs of two different people are digitally merged to create an image that sufficiently resembles both persons.
The image is then submitted as part of a legitimate passport application. If successful, this means that potentially, both persons could get away with using the same travel document.
A study by University of Lincoln in Lincoln, Britain, reveals that computers are more accurate than human beings at detecting such fraudulent identity photos. Psychologists at the university asked participants in an experiment to determine whether a photograph was of the person standing in front of them.
The study showed that about half the time, participants accepted the digitally-morphed images as the person in front of them. But, a computer programme was able to correctly identify the morphed images more than two-thirds of the time.
High-quality face morphs were used in the series of experiments designed to imitate real-life border control situations where an officer would have to accept or reject a passport image based on its resemblance to the person in front of them.
Participants in the experiment failed to spot 51% of the fraudulent images, and even when they were provided with information on face-morphing scams, the detection rates only rose to 64%.
Another experiment revealed that training did not help participants much in detecting the face morphed images presented on screen.
The results suggest that there is a risk that these fraudulent images will be accepted as legitimate photo identification, especially in border control situations where the validity of the travel document is often determined by a human officer.
When these images were put through a simple computer algorithm designed to differentiate between morphs and regular photos, 68% of the images were correctly identified as morphs. This reveals that the programme was significantly more accurate than the human participants.
The algorithm used was a relatively basic one just for demonstration purposes, but recent software developed is revealed to be far more sophisticated with even higher levels of accuracy.
Lead researcher Dr Robin Kramer said that the research could be of significance for security agencies and situations such as border control, where the use of computer algorithms might be a better way of minimising such frauds from slipping through the border.