一句话：提出基于模型变换与 delta debugging 数据增强的训练框架 Dare，显著提升深度学习模型的通用鲁棒性。
深度学习模型的安全性日益重要，提升其通用鲁棒性十分关键。现有方法中，对抗训练最具代表性，它通过加入对抗样本进行后训练调优，但面对多种攻击时仍存在泛化性不足、效果不理想的问题。本文提出一种新的模型训练框架，通过模型变换并结合 delta debugging 风格的数据增强策略，提升深度学习模型的通用鲁棒性。作者实现了工具 Dare，并在 9 个深度学习模型上进行了广泛评估。结果表明，Dare 显著优于现有对抗训练技术：在 45 个不同攻击的测试场景中，Dare 在 29 个场景取得最高经验鲁棒性，而最佳基线方法仅覆盖 5 个场景。
Deep learning (DL) techniques have attracted much attention in recent years, and have been applied to many application scenarios, including those that are safety-critical. Improving the universal robustness of DL models is vital and many approaches have been proposed in the last decades aiming at such a purpose. Among existing approaches, adversarial training is the most representative. It advocates a post model tuning process via incorporating adversarial samples. Although successful, they still suffer from the challenge of generalizability issues in the face of various attacks with unsatisfactory effectiveness. Targeting this problem, in this paper we propose a novel model training framework, which aims at improving the universal robustness of DL models via model transformation incorporated with a data augmentation strategy in a delta debugging fashion. We have implemented our approach in a tool, called Dare, and conducted an extensive evaluation on 9 DL models. The results show that our approach significantly outperforms existing adversarial training techniques. Specifically, Dare has achieved the highest Empirical Robustness in 29 of 45 testing scenarios under various attacks, while the number drops to 5 of 45 for the best baseline approach.
@inproceedings{paperbot7147,
title = {Toward Improving the Robustness of Deep Learning Models via Model Transformation},
author = {Yingyi Zhang and Zan Wang and Jiajun Jiang and Hanmo You and Junjie Chen},
booktitle = {ASE},
year = {2022},
doi = {10.1145/3551349.3556920}
}