Artificial Intelligence has advanced enough that it is now capable of fooling the biometric identity scanners by making fake fingerprints.
A research team of the New York University have established a neural network that could artificially make fingerprints and use the fake fingerprints for verification or authentication, as needed by the various biometric sensors used on the different tablets and smartphones, these days.
As per a report from The Guardian, the new Neural Network referred to as the “DeepMasterPrints” was able to copy fingerprints with very little error let’s say one out of thousand. Other than this a complete fingerprint was matched seventy-seven per cent of the time.
The new neural network was fed with a lot of real fingerprints initially. What the network did next was to apply the chunks of stored fingerprints with the verifier in a process very much similar to what is done under a dictionary password attack.
The research team applied the neural network to many fingerprints and were capable of generating replicas as much as twenty-three per cent of the total fingerprint that had to be made, having an error ratio of less than one in a thousand.
This percentage match is enough for bypassing the current fingerprint scanners which just looks for partial matches and permits one to pass if matched.
The commonly used fingerprint scanners do a comparison of the applied fingerprint impression to just one portion of the total fingerprint stored in the database.
Full fingerprints are super unique, the created fingerprint is capable of bypassing at least one portion of the fingerprint dataset stored on the verification device.