printer friendly version

Is a face a face or a fake?

Apple recently released its most ambitious product, the iPhone X, in celebration of the 10th anniversary of the release of the first iPhone. The iPhone X especially stands out with its introduction of Face ID, a face recognition technology to replace the fingerprint recognition (Touch ID).

According to Apple, Face ID will recognise a user’s face incorrectly only once in 1 million times, an immense improvement over the once in 50 000 times probability of Touch ID.

However, authentication through face recognition is actually as prone to spoofing as fingerprint-based methods. The press was quick to point out after Face ID was announced that it would be easy to recreate through 3D rendering the faces of well-known politicians or actors that are often exposed in photos or videos as there is a great amount of data available on them. 3D printers, which are used to build face structures, are also no longer hard to find. They can be commonly found in online shopping malls.

To prove such spoofing is possible, Bkav, a security company based in Vietnam succeeded in spoofing Face ID with a face mask it created. The mask’s face structure was created using a plastic mask and silicon, and an additional application of paper and makeup was enough to bypass authentication.

Then what about Suprema’s FaceStation 2, which also authenticates users with face recognition? Is its security also vulnerable to similar methods?

Advanced face recognition technology

FaceStation 2’s face recognition method functions differently from the iPhone X’s. The mask Bkav used to spoof Face ID has eyes, a nose, and a mouth constructed to resemble an actual person’s, but the remaining parts of the face are made of plastic. Through this information, it can be inferred that Face ID uses 3D modelling data and True Depth technology in its authentication process.

FaceStation 2, on the other hand, not only uses 140 near-infrared (NIR) LEDs to extract 3D modelling data, but also uses an algorithm to produce various numerical values from the infrared light reflected off the face. These values include those related to skin tone and skin texture values, which are helpful in distinguishing between similar faces. This special hardware and algorithm render the mask built by Bkav useless when tested against FaceStation 2’s authentication system.

FaceStation 2 also uses a matching score when executing face recognition. The matching score is a calculation of the match between a pre-registered face template and the actual face to be authenticated. The score helps guard FaceStation 2 against spoofing, as a low value is produced if the mask used in spoofing does not use a 3D model that exactly matches the actual face.

FaceStation 2 was recognised for its technological advancement at the 2017 Detektor International Awards, and was awarded the Best Product Award in the ID & Access category.

FaceStation 2 combines Suprema’s most advanced algorithm and hardware to provide a new contactless authentication experience, with a greater variety of features and increased performance.

Share this article:

Further reading: