AdjointDEIS: Efficient Gradients for Diffusion ModelsMila – Quebec AI InstituteAbstract. Training-free guided generation is a powerful technique to exert additional control on the generative pipeline of diffusion models. We discuss the application of the continuous adjoint equations on diffusion models. We then use these equations to estimate the gradients of the solution trajectory and conditional information with respect to some differentiable guiding function defined on the output. We discuss special optimizations for Variance Preserving (VP) type diffusions and demonstrate an application of this technique as an adversarial attack against a Face Recognition system in the form of a face morphing attack.
Diffusion Morphs: Leveraging Diffusion for Strong and High-Quality Face Morphing AttacksIEEE International Joint Conference on Biometrics — Journal TrackAbstract. Face morphing attacks seek to deceive a Face Recognition (FR) system by presenting a morphed image consisting of the biometric qualities from two different identities with the aim of triggering a false acceptance with one of the two identities. We present a novel morphing attack that uses a Diffusion-based architecture to improve the visual fidelity of the image and the ability of the morphing attack to represent characteristics from both identities. We demonstrate effectiveness via the Fréchet Inception Distance and extensive experiments measuring the vulnerability of FR systems.
2024 Jul 23
Diffusion Morphs: The Power of Iterative Generative Models for Attacking FR SystemsBiometrics Security & Privacy Group — Idiap Research InstituteAbstract. DiMs are a recently proposed morphing attack achieving SOTA performance for representation-based morphing. This talk covers three recent advancements: (1) Fast-DiM improves inference speed via higher-order ODE solvers; (2) Greedy-DiM increases FR vulnerability via a greedy optimization strategy during each step of generation; (3) AdjointDEIS, a novel strategy for backpropagating gradients through diffusion ODEs/SDEs via adjoint sensitivity.