Boffins craft perfect 'head generator' to beat facial recognition

Think Face/Off, in software, plus some digital touchup

Researchers from the Max Planck Institute for Informatics have defeated facial recognition on big social media platforms – by removing faces from photos and replacing them with automatically-painted replicas.

As the team of six researchers explained in their arXiv paper this month, people who want to stay private often blur their photos, not knowing that this is “surprisingly ineffective against state-of-the-art person recognisers.”

Inpainting example

AI can beat blurring, but not the inpainting adversarial model

The researchers went a step further by constructing a realistic “fake face” that replaced the original image, but was perturbed enough to beat the AI while still looking “right” to a friend perusing the image.

As outlined in the paper, the researchers had to “factor the head inpainting task into two stages: landmark detection/generation and head inpainting conditioned on body context and landmarks”.

The result, the boffins claimed, is that their model can provide a realistic-looking result, even when it's faced with “challenging poses and scenarios” including different lighting conditions, such that the “fake” face “blends naturally into the context”.

In common with modern facial recognition systems, Sun's software builds a point cloud of landmarks captured from someone's face; its adversarial attack against recognition perturbed those points.

Pairs of points from the original landmarks (real) and the generated landmarks (fake) are fed into the “head generator and discriminator” software to create the inpainted face. ?


Biting the hand that feeds IT ? 1998–2017

                                    1. 3239961348 2018-02-21
                                    2. 8189611347 2018-02-21
                                    3. 1166571346 2018-02-21
                                    4. 905911345 2018-02-21
                                    5. 238301344 2018-02-21
                                    6. 9856121343 2018-02-21
                                    7. 7107891342 2018-02-21
                                    8. 616201341 2018-02-21
                                    9. 97671340 2018-02-21
                                    10. 7844621339 2018-02-21
                                    11. 9607131338 2018-02-21
                                    12. 3095441337 2018-02-21
                                    13. 9602111336 2018-02-21
                                    14. 5723751335 2018-02-21
                                    15. 1275371334 2018-02-21
                                    16. 8517591333 2018-02-21
                                    17. 230661332 2018-02-21
                                    18. 3311101331 2018-02-21
                                    19. 6181321330 2018-02-20
                                    20. 6139401329 2018-02-20