Quantifying usability in secure graphics

There is an increasing need for methods for secure dissemination of interactive 3D graphics content, providing protection for valuable 3D models while still allowing them to be widely shared. Existing systems for protected sharing of 3D models may introduce perturbations into the rendered images of the content, in order to defend against potential malicious reconstruction attacks that could otherwise recover the 3D model shape. However, the particular nature and magnitude of these perturbation defenses has not been based upon any rigorous analysis or measurement of their perceptual effect on non-malicious users of the protected graphics system. In this paper, we take the first steps toward such an analysis, conducting a series of user studies that evaluate the impact (as measured by user reaction time) of varying amounts of noise applied to user interactions in a real-time 3D rendering system. We are thus able to experimentally determine the most appropriate tradeoffs between noise perturbation defenses and the security of the 3D content against typical reconstruction attacks.