(Certified!!) Adversarial Robustness for Free!

Nicholas Carlini*, Florian Tramèr*, Rice Leslie, Mingjie Sun, Krishnamurthy Dvijotham and J Zico Kolter   (*joint first authors)

International Conference on Learning Representations (ICLR) 2023


In this paper we show how to achieve state-of-the-art certified adversarial robustness to 2-norm bounded perturbations by relying exclusively on off-the-shelf pretrained models. To do so, we instantiate the denoised smoothing approach of Salman et al. by combining a pretrained denoising diffusion probabilistic model and a standard high-accuracy classifier. This allows us to certify 71% accuracy on ImageNet under adversarial perturbations constrained to be within a 2-norm of 0.5, an improvement of 14 percentage points over the prior certified SoTA using any approach, or an improvement of 30 percentage points over denoised smoothing. We obtain these results using only pretrained diffusion models and image classifiers, without requiring any fine tuning or retraining of model parameters.

  author   =   {Carlini, Nicholas and Tram{\`e}r, Florian and Leslie, Rice and Sun, Mingjie and Dvijotham, Krishnamurthy and Kolter, J Zico},
  title   =   {(Certified!!) Adversarial Robustness for Free!},
  booktitle   =   {International Conference on Learning Representations (ICLR)},
  year   =   {2023},
  howpublished   =   {arXiv preprint arXiv:2206.10550},
  url   =   {https://arxiv.org/abs/2206.10550}