As Firm As Their Foundations: Can open-sourced foundation models be used to create adversarial examples for downstream tasks?

arXiv:2403.12693v1 Announce Type: new
Abstract: Foundation models pre-trained on web-scale vision-language data, such as CLIP, are widely used as cornerstones of powerful machine learning systems. While pre-training offers clear advantages for downstream learning, it also endows downstream models with shared adversarial vulnerabilities that can be easily identified through the open-sourced foundation model. In this work, we expose such vulnerabilities in CLIP's downstream models and show that foundation models can serve as a basis for attacking their downstream systems. In particular, …

abstract advantages adversarial adversarial examples arxiv clear clip cs.cv data examples foundation language language data learning systems machine machine learning pre-training scale systems tasks training type vision vulnerabilities web