【论文阅读】APMSA: Adversarial Perturbation Against Model Stealing Attacks(2023)
摘要TrainingaDeepLearning(DL)model(训练深度学习模型)requiresproprietarydata(专有数据)andcomputing-intensiveresources(计算密集型资源).Torecouptheirtrainingcosts(收回训练成本),amodelprovidercanmonetizeDLmodelsthroughMachineLearni