MultiRobustBench
收藏资源简介:
MultiRobustBench是由普林斯顿大学电气与计算机工程系的研究团队开发的一个用于评估机器学习模型对抗多种攻击的鲁棒性的基准数据集。该数据集包含180种不同的攻击类型,旨在为机器学习模型提供一个全面的测试平台,以评估其在面对未知攻击时的表现。数据集通过一个统一的框架来模拟不同程度的对抗知识,允许研究者评估模型在面对联合攻击和未知攻击时的鲁棒性。MultiRobustBench的访问地址为https://multirobustbench.github.io,是当前最大的多攻击评估数据集之一,对于推动机器学习模型在安全关键应用中的鲁棒性研究具有重要意义。
MultiRobustBench is a benchmark dataset developed by the research team from the Department of Electrical and Computer Engineering at Princeton University, intended to evaluate the robustness of machine learning models against a wide range of adversarial attacks. This dataset comprises 180 distinct attack types, providing a comprehensive testbed for machine learning models to assess their performance when confronted with unseen attacks. Leveraging a unified framework, the dataset simulates varying levels of adversarial knowledge, allowing researchers to evaluate model robustness against both combined and unknown adversarial attacks. The official access link for MultiRobustBench is https://multirobustbench.github.io. As one of the largest multi-attack evaluation datasets currently available, MultiRobustBench plays a critical role in advancing robustness research for machine learning models in safety-critical applications.




