Artificial Intelligence manages to deceive readers of fingerprints
An Artificial Intelligence (AI) system based on machine learning has been able to produce a fingerprint that mimics the fingerprints of one out of every five people, and can deceive the biometric recognition sensors of electronic devices.
The system comes from an academic study, headed by the Tandon engineering school of the University of New York (United States) and in which the State University of Michigan has also participated, has alerted about the vulnerability in the biometric recognition mechanisms in current devices, such as smartphones.
Researchers have used a neural network that synthesizes human fingerprints. To do this, it uses an AI algorithm that takes as an example an open fingerprint database, and imitates distinctive aspects of them. Fingerprints work like a master key that is able to deceive footprint readers that are part of biometric recognition, with potential effectiveness for one in five people.
The study is based on the fact that biometric recognition sensors use partial fingerprints to confirm the identity of the user. These fingerprints are less unique than the complete ones and the system takes advantage of this to falsify them. This type of experiments “demonstrate the need for a multi-factor authentication”, as Philip Bontrager, the main author of the research, has asserted, in reference to the current impossibility of demonstrating whether the fingerprints presented to a sensor come from a human being.
Source: ABC Software