模型反转攻击样本数据集
收藏资源简介:
本研究创建了一个包含28个不同模型反转攻击、防御设置的数据集,基于私人和公开数据集。数据集包含了多种模型和攻击方式,旨在评估模型反转攻击的准确性和有效性。数据集的创建过程涉及了多种模型反转攻击的设置和防御策略,并通过人工标注的方式确保了数据集的质量。该数据集可用于评估模型反转攻击的准确性和有效性,以及开发更可靠和鲁棒的模型反转攻击防御方法。
This study constructs a dataset containing 28 distinct model inversion attack and defense configurations, built upon both private and public datasets. The dataset incorporates a wide range of models and attack methods, with the primary objective of evaluating the accuracy and effectiveness of model inversion attacks. The dataset construction process involves various model inversion attack setups and defense strategies, while its quality is guaranteed through manual annotation. This dataset can serve as a platform for evaluating the accuracy and effectiveness of model inversion attacks, as well as for developing more reliable and robust defense methods against such attacks.

- 1Uncovering the Limitations of Model Inversion Evaluation: Benchmarks and Connection to Type-I Adversarial Attacks新加坡科技设计大学(SUTD)和马格德堡奥托·冯·格里克大学 · 2025年



