Deceptive Diffusion: Generating Synthetic Adversarial Examples
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909233036918784 |
|---|---|
| author | Beerens, Lucas Higham, Catherine F. Higham, Desmond J. |
| author_facet | Beerens, Lucas Higham, Catherine F. Higham, Desmond J. |
| contents | We introduce the concept of deceptive diffusion -- training a generative AI model to produce adversarial images. Whereas a traditional adversarial attack algorithm aims to perturb an existing image to induce a misclassificaton, the deceptive diffusion model can create an arbitrary number of new, misclassified images that are not directly associated with training or test images. Deceptive diffusion offers the possibility of strengthening defence algorithms by providing adversarial training data at scale, including types of misclassification that are otherwise difficult to find. In our experiments, we also investigate the effect of training on a partially attacked data set. This highlights a new type of vulnerability for generative diffusion models: if an attacker is able to stealthily poison a portion of the training data, then the resulting diffusion model will generate a similar proportion of misleading outputs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_19807 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Deceptive Diffusion: Generating Synthetic Adversarial Examples Beerens, Lucas Higham, Catherine F. Higham, Desmond J. Machine Learning Artificial Intelligence 68T07 I.2.0; I.5.1 We introduce the concept of deceptive diffusion -- training a generative AI model to produce adversarial images. Whereas a traditional adversarial attack algorithm aims to perturb an existing image to induce a misclassificaton, the deceptive diffusion model can create an arbitrary number of new, misclassified images that are not directly associated with training or test images. Deceptive diffusion offers the possibility of strengthening defence algorithms by providing adversarial training data at scale, including types of misclassification that are otherwise difficult to find. In our experiments, we also investigate the effect of training on a partially attacked data set. This highlights a new type of vulnerability for generative diffusion models: if an attacker is able to stealthily poison a portion of the training data, then the resulting diffusion model will generate a similar proportion of misleading outputs. |
| title | Deceptive Diffusion: Generating Synthetic Adversarial Examples |
| topic | Machine Learning Artificial Intelligence 68T07 I.2.0; I.5.1 |
| url | https://arxiv.org/abs/2406.19807 |