Neural Network Learned to Fool Face Identification Systems With “Universal” Face

Neural Network Learned to Fool Face Identification Systems With “Universal” Face

Scientists from Tel Aviv University in Israel have developed a way to bypass biometric identification systems with a fake facial image.

The discovered method uses artificial intelligence technologies to create a universal facial template that can consistently combine and unlock identity verification systems.

According to the researchers, the vulnerability lies in the fact that facial recognition uses a broad set of tokens to identify specific people. These can be used to create a universal template that can fool a large percentage of security systems.

Scientists suggested that as few as 9 faces generated by the StyleGAN algorithm could match 40 percent of the world’s population. They tested the neural network’s synthesized snapshot on a dataset of 13,000 images and found that the fake face could mimic 20 percent of the dataset’s personalities. Other tests showed better results.

“Face authentication is extremely vulnerable, even if attackers have no information about the target identity,” the researchers reported.

Researchers also believe the vulnerability could be combined with dipfakes to mimic the “liveliness” verification often used to remotely confirm identities by performing a series of gestures.

Recall that in July, researchers reported discovering a method to stealthily inject malicious code into neural networks.

In May, researchers from the Maryland Cybersecurity Center identified a vulnerability in neural networks that increases their power consumption.

Hughie Marsh

About Hughie Marsh

HI, guys I am Hughie Marsh Me and Team are here to share knowledge about top useful applications and all the tech news around the world. Stay tuned and download our monthly Digital Issues!