Improving Adversarial Transferability via Model Alignment
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2023
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910529882161152 |
|---|---|
| author | Ma, Avery Farahmand, Amir-massoud Pan, Yangchen Torr, Philip Gu, Jindong |
| author_facet | Ma, Avery Farahmand, Amir-massoud Pan, Yangchen Torr, Philip Gu, Jindong |
| contents | Neural networks are susceptible to adversarial perturbations that are transferable across different models. In this paper, we introduce a novel model alignment technique aimed at improving a given source model's ability in generating transferable adversarial perturbations. During the alignment process, the parameters of the source model are fine-tuned to minimize an alignment loss. This loss measures the divergence in the predictions between the source model and another, independently trained model, referred to as the witness model. To understand the effect of model alignment, we conduct a geometric analysis of the resulting changes in the loss landscape. Extensive experiments on the ImageNet dataset, using a variety of model architectures, demonstrate that perturbations generated from aligned source models exhibit significantly higher transferability than those from the original source model. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2311_18495 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Improving Adversarial Transferability via Model Alignment Ma, Avery Farahmand, Amir-massoud Pan, Yangchen Torr, Philip Gu, Jindong Machine Learning Computer Vision and Pattern Recognition Neural networks are susceptible to adversarial perturbations that are transferable across different models. In this paper, we introduce a novel model alignment technique aimed at improving a given source model's ability in generating transferable adversarial perturbations. During the alignment process, the parameters of the source model are fine-tuned to minimize an alignment loss. This loss measures the divergence in the predictions between the source model and another, independently trained model, referred to as the witness model. To understand the effect of model alignment, we conduct a geometric analysis of the resulting changes in the loss landscape. Extensive experiments on the ImageNet dataset, using a variety of model architectures, demonstrate that perturbations generated from aligned source models exhibit significantly higher transferability than those from the original source model. |
| title | Improving Adversarial Transferability via Model Alignment |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2311.18495 |