Joint work with @parisgiampouras, @ReneVid38572141 at JHU. We tackle the problem of provable joint signal and attack classification for adversarial examples by posing as a block-sparse recovery problem.

If you're at ICML, come check our poster/talk on "Reverse Engineering $\ell_p$ attacks: A block-sparse optimization approach with recovery guarantees"! Poster: Room 327-329 Tuesday, 6:30pm-8:30pm Presentation: Room 327-329 Tuesday, 11am Paper: https://t.co/wx3rqi3fwm #ICML2022