TSS: Transformation-Specific Smoothing for Robustness Certification

Linyi Li, Maurice Weber, Xiaojun Xu, Luka Rimanic, Bhavya Kailkhura, Tao Xie, Ce Zhang, Bo Li

Abstract: As machine learning (ML) systems become pervasive, safeguarding their security is critical. However, recently it has been demonstrated that motivated adversaries are able to mislead ML systems by perturbing test data using semantic transformations. While there exists a rich body of research providing provable robustness guarantees for ML models against Lp norm bounded adversarial perturbations, guarantees against semantic perturbations remain largely underexplored. In this paper, we provide TSS --- a unified framework for certifying ML robustness against general adversarial semantic transformations. First, depending on the properties of each transformation, we divide common transformations into two categories, namely resolvable (e.g., Gaussian blur) and differentially resolvable (e.g., rotation) transformations. For the former, we propose transformation-specific randomized smoothing strategies and obtain strong robustness certification. The latter category covers transformations that involve interpolation errors, and we propose a novel approach based on stratified sampling to certify the robustness. We develop the unified framework, including these certification strategies, with consistency-enhanced training approaches for rigorous robustness certification. We conduct extensive experiments on over ten types of challenging semantic transformations and show that our framework TSS significantly outperforms the state of the art. Moreover, to the best of our knowledge, TSS is the first that achieves nontrivial certified robustness on the large-scale ImageNet dataset. For instance, our framework achieves 30.4% certified robust accuracy against rotation attack (within 30 degrees) on ImageNet. Moreover, to consider a broader range of transformations, we show TSS is also robust against adaptive attacks and unforeseen image corruptions such as CIFAR-10-C and ImageNet-C.