CUDA

CUDA是由马里兰大学开发的基于卷积的不可学习数据集，旨在通过添加特定设计的噪声来保护数据隐私。该数据集通过控制类别的卷积操作，使用随机生成的滤波器对训练数据进行处理，使得模型学习到的是滤波器与标签之间的关系而非数据的有用特征。CUDA适用于解决深度学习模型在未经授权使用在线数据时引发的数据隐私问题，通过降低清洁测试数据的准确性来实现其效果。

CUDA is a convolution-based non-learnable dataset developed by the University of Maryland, which aims to protect data privacy by adding specially designed noise. This dataset processes training data using randomly generated filters via category-controlled convolution operations, enabling models to learn the relationship between filters and labels rather than the useful features inherent in the original data. CUDA is designed to address data privacy issues caused by unauthorized use of online data by deep learning models, and achieves its protective effect by reducing the accuracy of clean test data.