Deceptive Diffusion: Generating Synthetic Adversarial Examples
This addresses the challenge of robust AI defense for security applications by providing a scalable source of adversarial examples, though it is incremental as it builds on existing diffusion models.
The paper tackles the problem of generating adversarial examples by introducing deceptive diffusion, a method to train generative AI models to produce misclassified images, enabling scalable adversarial training data. The result includes demonstrating a vulnerability where poisoning part of the training data leads to a proportional rate of misleading outputs.
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.