Don't miss the latest stories
AI Successfully Creates ‘Universal Fingerprints’ That Can Unlock Smartphones
By Mikelle Leow, 16 Nov 2018
Image via Shutterstock
Aside from being accurate, biometric authentications bring about a kind of flexibility that old-fashioned passwords lack. Whereas passwords are either right or wrong, a sensor that reads faces, fingerprints, or irises is more likely to approve an entry even when a subject has had a haircut.
This versatility might be an advantage, but here’s the frightening bit: an AI has figured out how to exploit it.
Researchers at New York University and the University of Michigan have published a paper detailing how they have taught a machine learning algorithm to fashion artificial “fingerprints,” which they call ‘DeepMasterPrints’, that serve as “master keys” to a “large number” of biometric scanners.
To train the system, the team fed it with the fingerprints of over 6,000 people until it was able to create its own. Its final products were pit against the same algorithms present in the fingerprint scanners of phones; if the prints were rejected, the AI would tweak the designs until they went through.
It’s worth noting that the “fingerprints” do not have to match real ones. After the AI’s in-depth analysis, the researchers were able to generate images containing the most common fingerprint patterns that could trick most scanners into giving false positives.
While the system isn’t perfect, it’s already able to pass through scanners that grant access to one-percent false positives with a 77-percent success rate.
With stricter fingerprint scanners with 0.1-percent unauthorized entries, the AI is able to get through 22-percent of the time. Further, ‘DeepMasterPrints’ have a one-percent success rate at passing scanners that are known to allow only 0.01% of false positives through.
The researchers describe that the test was designed to prove “the vulnerability of fingerprint recognition systems.” Perhaps this sort of weakness was what spurred Apple to ditch Touch ID in iPhones for facial recognition.
Real fingerprints (left) VS ‘DeepMasterPrints’ (right). Screenshot via DeepMasterPrints
[via Android Police, images via various sources]
More related news
Also check out these recent news