The field of machine learning is witnessing significant developments in adversarial attacks and generative models. Researchers are proposing novel methods to improve the transferability of adversarial attacks, such as using background mixup and temporal consistency constraints. Additionally, there is a growing interest in leveraging diffusion models for synthetic data augmentation, which has shown promising results in improving model robustness. Furthermore, studies are investigating the mechanisms underlying the improvements in adversarial robustness achieved by diffusion models, highlighting the importance of compression effects and internal randomness. Noteworthy papers in this area include those that propose innovative approaches to adversarial training, such as using energy-based models and delta energy regularizers, as well as those that develop new generative models for specific applications, like cryo-electron microscopy synthesis.
Advances in Adversarial Attacks and Generative Models
Sources
Ownership Verification of DNN Models Using White-Box Adversarial Attacks with Specified Probability Manipulation
Temporal Consistency Constrained Transferable Adversarial Attacks with Background Mixup for Action Recognition
Towards more transferable adversarial attack in black-box manner
F-ANcGAN: An Attention-Enhanced Cycle Consistent Generative Adversarial Architecture for Synthetic Image Generation of Nanoparticles
Do We Need All the Synthetic Data? Towards Targeted Synthetic Image Augmentation via Diffusion Models
Modeling extreme events and intermittency in turbulent diffusion with a mean gradient
What is Adversarial Training for Diffusion Models?
Rethinking Gradient-based Adversarial Attacks on Point Cloud Classification
Understanding Adversarial Training with Energy-based Models
How Do Diffusion Models Improve Adversarial Robustness?
Leveraging Diffusion Models for Synthetic Data Augmentation in Protein Subcellular Localization Classification
CryoCCD: Conditional Cycle-consistent Diffusion with Biophysical Modeling for Cryo-EM Synthesis