On Adversarial Robustness of Deep Image Deblurring
This highlights a security risk for image recovery systems, though it is incremental as it extends adversarial robustness from classification to a new domain.
The paper tackles the vulnerability of deep learning-based image deblurring methods to adversarial attacks, showing that imperceptible distortions can significantly degrade performance and alter output content.
Recent approaches employ deep learning-based solutions for the recovery of a sharp image from its blurry observation. This paper introduces adversarial attacks against deep learning-based image deblurring methods and evaluates the robustness of these neural networks to untargeted and targeted attacks. We demonstrate that imperceptible distortion can significantly degrade the performance of state-of-the-art deblurring networks, even producing drastically different content in the output, indicating the strong need to include adversarially robust training not only in classification but also for image recovery.