Adversarial Illusions in Multi-Modal Embeddings

Authors: 

Tingwei Zhang and Rishi Jha, Cornell University; Eugene Bagdasaryan, University of Massachusetts Amherst; Vitaly Shmatikov, Cornell Tech

Distinguished Paper Award Winner

Abstract: 

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 paper, 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.

These attacks are cross-modal and targeted: the adversary can align any image or sound with any target of his choice. Adversarial illusions exploit proximity in the embedding space and are thus agnostic to downstream tasks and modalities, enabling a wholesale compromise of current and future tasks, as well as modalities not available to the adversary. Using ImageBind and AudioCLIP embeddings, we demonstrate how adversarially aligned inputs, generated without knowledge of specific downstream tasks, mislead image generation, text generation, zero-shot classification, and audio retrieval.

We investigate transferability of illusions across different embeddings and develop a black-box version of our method that we use to demonstrate the first adversarial alignment attack on Amazon's commercial, proprietary Titan embedding. Finally, we analyze countermeasures and evasion attacks.

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.

BibTeX
@inproceedings {299866,
author = {Eugene Bagdasaryan and Rishi Jha and Vitaly Shmatikov and Tingwei Zhang},
title = {Adversarial Illusions in {Multi-Modal} Embeddings},
booktitle = {33rd USENIX Security Symposium (USENIX Security 24)},
year = {2024},
isbn = {978-1-939133-44-1},
address = {Philadelphia, PA},
pages = {3009--3025},
url = {https://www.usenix.org/conference/usenixsecurity24/presentation/zhang-tingwei},
publisher = {USENIX Association},
month = aug
}