This Quick Look provides a short preview of the full session presented at RSAC™ 2025 Conference, now available on demand. Multi-modal embeddings encode texts, images, thermal images, sounds, and videos into a single embedding space, aligning representations across different modalities (e.g., associate an image of a dog with a barking sound). In this presentation, we show that multi-modal embeddings can be vulnerable to an attack we call "adversarial illusions." Given an image or a sound, an adversary can perturb it to make its embedding close to an arbitrary, adversary-chosen input in another modality.

See the full session here.