id	author	title	date	pages	extension	mime	words	sentence	flesch	summary	cache	txt
easat-4184	Huang, Weiwang ; Hsieh, Ming Hsun 	The application of artificial intelligence and machine learning in civil law protection of privacy rights	2025	31	.pdf	application/pdf	15741	978	43	Adversary goals Adversary knowledge Adversary capabilities Adversary strategies Training data privacy, model privacy, prediction output privacy White-box: full model knowledge Strong: participate in training or access data Model inversion, extraction, and membership inference attacks Black-box: no model knowledge Weak: access limited outputs Adversarial attacks on machine learning (ML) models aim to compromise the confidentiality of these systems, focusing on accessing sensitive information such as training data privacy, model parameters, and prediction outcomes. In ML, 567 Edelweiss Applied Science and Technology ISSN: 2576-8484 Vol. 9, No. 1: 564-594, 2025 DOI: 10.55214/25768484.v9i1.4184 © 2025 by the author; licensee Learning Gate privacy can be categorised into three primary aspects: training data privacy, model privacy, and prediction output privacy.	cache/easat-4184.pdf	txt/easat-4184.txt