How secure is facial recognition?

Next story
Olivia Storey

Facial recognition software has been bought to the fore recently with the announcement of the iPhone X and its use by law enforcement, but how secure is it?

The algorithms within facial recognition software identify and measure the geometry of facial features; including the size and shape of eyes, nose, cheekbones and the jaw.

Newer 3D technology has a higher accuracy level than the original 2D technology, and is not affected by changes in lighting, viewing angles, different facial expressions, make up, or head placement.

After identifying and measuring these features, they are then compressed into data, only saving the parts of data useful to make a comparison.

In the UK facial recognition was used at the UEFA championships back in June. Faces were scanned in the Principality stadium and Cardiff’s central railway station to compare with and match against 500,000 ‘custody images’ stored by the local police.

The most recent and prominent use of facial recognition is the new iPhone X which uses facial recognition via infrared and 3D sensors within the two front facing cameras. It will be used to unlock the phone, authenticate logins and even Apple Pay.

Ondrej Kubovič, ESET Security Awareness Specialist, talks about the accuracy and safety of facial recognition.

“As far as safety aspects go, this technology should not endanger one’s health: in worst case it invades his/her privacy.

“Although, when it comes to accuracy that is a whole other story.

“Face recognition solutions are based on machine learning algorithms, which have limits that have been observed over the years of their implementation in various fields.

“In an environment that isn’t trying to evade its functionality, such as regular users and their pictures on social networks, machine learning can be pretty successful in identifying individuals.

“However in an adversarial environment, there are multiple ways how a subject can avoid the mechanisms of such technology.

“Reaching from basic ‘camouflage’ such as glasses, masks, face covers to more sophisticated and financially demanding approaches.

“Let’s say plastic surgery or creating masks with specialised pattern to fool the algorithm or look like other specific individual.

“And criminals will definitely do everything possible to stay under the radar.

“With machine learning there’s also always the question of false alarms, which occur rather frequently and thus require the algorithm to be supervised by humans.”

What do you think of facial recognition? Do you like the look of the iPhone X? Let us know on Twitter @ESETUK.

Join the ESET UK LinkedIn Group and stay up to date with the blog. If you are interested in seeing where ESET has been featured in the news then check out our ‘In the news’ section.