HOPE 2020 (2020): "Borders and Biometrics: Boundaries of Computer of Vision" (Download)
Sunday, August 2, 2020: 1900. Machine learning has been rapidly adopted by law enforcement as a way to justify sentencing, policing focus, and border control. In particular, facial recognition technology has been deployed around the world through massive surveillance networks, public/private 'fusion centers,' and cloud tools that make identifying an individual cheaper than a subway ride. The coming years will see similar technologies deployed in autonomous cars and drones despite research showing commercial systems disproportionately mis-classify non-males and people of color. These systems are mathematically opaque and rarely open source anyway, meaning that these algorithms govern our lives without corresponding oversight. Unsurprisingly, recent research suggests these systems are also fragile to many adversarial attacks. This talk will explore the history of biometric AI, its modern deployment, and fun ways to break these systems.