Camera Feature Attacks

Cameras have features like automatic exposure controls, auto-focus, and light-sensitivity. These can be targeted in attacks. Cameras normalize lighting conditions iteratively. Directing light at the image sensor can cause the camera to tune down its sensitivity and exposure, leading to undesired effects. For instance, auto exposure tuning down due to headlights at night could hide information in the background, such as traffic signs or pedestrians. The Google Driverless Car has been noted to be susceptible to this problem. These attacks are aiming to influence the camera's auto controls in the period before the image recovers and stabilizes. It's harder to detect because it consists of bursts of light instead of a constant beam. The longer it takes for the image to stabilize, the longer the car is vulnerable to undetected objects. This attack is different from situations like driving out of a tunnel, where the camera can adapt more gradually to new conditions.