Machine learning models robust to adversarial attacks

Abstract

В данной работе были рассмотрены современные подходы к состязательным атакам и защите от них, а так же представлены собственные решения задачи поиска атакованных объектов. Полученные методы были протестированы на известных наборах данных для популярных атак.
In this paper were considered modern approaches to adversarial attacks and defenses against them. There were also presented own solutions to the problem of searching for adversarial attacks. The obtained methods were tested on well-known data sets for popular attacks.

Description

Citation

Collections

Endorsement

Review

Supplemented By

Referenced By