- [2308. 11804] Adversarial Illusions in Multi-Modal Embeddings
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
- Adversarial Illusions in Multi-Modal Embeddings - USENIX
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
- Adversarial Illusions in Multi-Modal Embeddings - GitHub
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
- Rishi Jha
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
- USENIX Security ’24 Artifact Appendix: Adversarial Illusions in Multi . . .
The artifact demonstrates the vulner-ability of multi-modal embeddings to 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
- RSAC™ 2025 Conference Quick Look: Adversarial Illusions in Multi-Modal . . .
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
- [2511. 21893] Breaking the Illusion: Consensus-Based Generative . . .
Multi-modal foundation models align images, text, and other modalities in a shared embedding space but remain vulnerable to adversarial illusions (Zhang et al , 2025), where imperceptible perturbations disrupt cross-modal alignment and mislead downstream tasks To counteract the effects of adversarial illusions, we propose a task-agnostic mitigation mechanism that reconstructs the input from
- AUTOENCODIX: a generalized and versatile framework to train and . . .
These architectures, such as ontology-based and cross-modal autoencoders, provide key advantages over traditional methods by offering explainability of embeddings or the ability to translate
|