|
ИСТИНА |
Войти в систему Регистрация |
ФНКЦ РР |
||
CC-CERT: A Probabilistic Approach to Certify General Robustness of Neural Networksстатья Исследовательская статья
- Авторы: Pautov Mikhail, Tursynbek Nurislam, Munkhoeva Marina, Muravev Nikita, Petiushko Aleksandr, Oseledets Ivan
- Журнал: Proceedings of the ... AAAI Conference on Artificial Intelligence. AAAI Conference on Artificial Intelligence
- Том: 36
- Номер: 7
- Год издания: 2022
- Издательство: AAAI Press
- Местоположение издательства: Menlo Park, Calif., United States
- Первая страница: 7975
- Последняя страница: 7983
- DOI: 10.1609/aaai.v36i7.20768
- Аннотация: In safety-critical machine learning applications, it is crucial to defend models against adversarial attacks — small modifications of the input that change the predictions. Besides rigorously studied $\ell_p$-bounded additive perturbations, semantic perturbations (e.g. rotation, translation) raise a serious concern on deploying ML systems in real-world. Therefore, it is important to provide provable guarantees for deep learning models against semantically meaningful input transformations. In this paper, we propose a new universal probabilistic certification approach based on Chernoff-Cramer bounds that can be used in general attack settings. We estimate the probability of a model to fail if the attack is sampled from a certain distribution. Our theoretical findings are supported by experimental results on different datasets.
- Добавил в систему: Муравьев Никита Валерьевич