New method makes private AI training more secure and efficient

Original: Dithered Gaussian Mechanism for Randomness-Efficient Differential Privacy

Source: arxiv.org ↗