Latent Transfer Attack: Adversarial Examples via Generative Latent Spaces

Abstract

Latent Transfer Attack optimizes image perturbations in the latent space of a pretrained generative model. Transformations and latent smoothing help study how these perturbations transfer across architectures and preprocessing pipelines under a pixel-space constraint.

Ravid Shwartz-Ziv

AI Researcher

AI researcher at Meta MSL with a background in information theory and computational neuroscience, working on world models, memory, and compression. Former Assistant Professor and Faculty Fellow at NYU, collaborating on research across academia and industry.