CQT Online Talks – Series: Computer Science Seminars (live)
Speaker: Christoph Hirche, CQT
Abstract: Differential privacy has been an exceptionally successful concept when itcomes to providing provable security guarantees for classical computations.More recently, the concept was generalized to quantum computations. Whileclassical computations are essentially noiseless and differential privacy isoften achieved by artificially adding noise, near-term quantum computers areinherently noisy and it was observed that this leads to natural differentialprivacy as a feature. In this work we discuss quantum differential privacy inan information theoretic framework by casting it as a quantum divergence. Amain advantage of this approach is that differential privacy becomes a propertysolely based on the output states of the computation, without the need to checkit for every measurement. This leads to simpler proofs and generalizedstatements of its properties as well as several new bounds for both, generaland specific, noise models. In particular, these include common representationsof quantum circuits and quantum machine learning concepts. Here, we focus onthe difference in the amount of noise required to achieve certain levels ofdifferential privacy versus the amount that would make any computation useless.Finally, we also generalize the classical concepts of local differentialprivacy, Rényi differential privacy and the hypothesis testing interpretationto the quantum setting, providing several new properties and insights.
Смотрите видео Quantum Differential Privacy: An Information Theory Perspective онлайн без регистрации, длительностью часов минут секунд в хорошем качестве. Это видео добавил пользователь Centre for Quantum Technologies 27 Апрель 2022, не забудьте поделиться им ссылкой с друзьями и знакомыми, на нашем сайте его посмотрели 445 раз и оно понравилось 6 людям.