Robust Unpaired Single Image Super-Resolution of Faces
This addresses a specific vulnerability in facial image processing for security and quality applications, but it is incremental as it builds on existing adversarial attack frameworks.
The paper tackles the problem of adversarial attacks on facial super-resolution networks by proposing a method that identifies a parameterizable property of the MSE loss to locate optimal degradations efficiently, achieving a better speed-effectiveness trade-off than state-of-the-art attacks like FGSM and PGD.
We propose an adversarial attack for facial class-specific Single Image Super-Resolution (SISR) methods. Existing attacks, such as the Fast Gradient Sign Method (FGSM) or the Projected Gradient Descent (PGD) method, are either fast but ineffective, or effective but prohibitively slow on these networks. By closely inspecting the surface that the MSE loss, used to train such networks, traces under varying degradations, we were able to identify its parameterizable property. We leverage this property to propose an adverasrial attack that is able to locate the optimum degradation (effective) without needing multiple gradient-ascent steps (fast). Our experiments show that the proposed method is able to achieve a better speed vs effectiveness trade-off than the state-of-theart adversarial attacks, such as FGSM and PGD, for the task of unpaired facial as well as class-specific SISR.