Facial recognition systems can be vulnerable to deep morphing, researchers say

Some 95 percent of deepfakes are accepted by biometric facial recognition systems, found a study conducted by Pavel Korshunov and Sebastien Marcel from Idiap Research Institute in Martigny, Switzerland.

According to Korshunov and Marcel, current facial recognition systems are vulnerable to high-quality fake images and videos created using generative adversarial networks (GANs), creating a need for automated detection of GAN-generated faces. They used open source software based on GANs to create deepfake videos with faces morphed with a GAN-based algorithm, to prove that “state of the art face recognition systems based on VGG and Facenet neural networks are vulnerable to the deep morph videos, with 85.62 percent and 95.00 percent false acceptance rates, respectively, which means methods for detecting these videos are necessary.”

They found that visual quality metrics are most effective in detecting deep morphs with 8.97 percent equal error rate. The study is called Vulnerability of Face Recognition to Deep Morphing and can be reviewed here. The research were discussed at the Frontex International Conference on Biometrics for Borders 2019 in Warsaw.

Google, in partnership with Jigsaw, recently produced and delivered a massive database of visual deepfakes that is now part of the FaceForensics benchmark created by the Technical University of Munich and the University Federico II of Naples. The database has hundreds of recorded videos which were manipulated with widely available deepfake generation methods to create thousands of deepfakes.

Other research from Amsterdam-based cybersecurity company Deeptrace warns that deepfakes are spreading extremely fast online, “with the number of deepfake videos almost doubling over the last seven months to 14,678.” This is possible thanks to a high number of commodification tools that make it easier for individuals to create deepfakes and disseminate them through social media. The company noticed a high number of deepfakes and synthetic media tools arising from China and South Korea.

Article Topics

biometrics  |  deepfakes  |  face  |  face morphing  |  facial recognition  |  Idiap  |  spoof detection