Black-Box Certification with Randomized Smoothing: A Functional Optimization Based Framework

Authors: Dinghuai Zhang, Mao Ye, Chengyue Gong, Zhanxing Zhu, Qiang Liu
Published: 2020-02-21 | Updated: 2020-10-20

Source: https://arxiv.org/abs/2002.09169

PDF: https://arxiv.org/pdf/2002.09169

AIにより推定されたラベル

ロバスト性評価 最適化問題 防御手法

※ こちらのラベルはAIによって自動的に追加されました。そのため、正確でないことがあります。
詳細は文献データベースについてをご覧ください。

Abstract

Randomized classifiers have been shown to provide a promising approach for achieving certified robustness against adversarial attacks in deep learning. However, most existing methods only leverage Gaussian smoothing noise and only work for ℓ₂ perturbation. We propose a general framework of adversarial certification with non-Gaussian noise and for more general types of attacks, from a unified functional optimization perspective. Our new framework allows us to identify a key trade-off between accuracy and robustness via designing smoothing distributions, helping to design new families of non-Gaussian smoothing distributions that work more efficiently for different ℓ_p settings, including ℓ₁, ℓ₂ and ℓ_∞ attacks. Our proposed methods achieve better certification results than previous works and provide a new perspective on randomized smoothing certification.