Adversarial Examples in Deep Learning for Multivariate Time Series Regression

Research on Adversarial Attacks on Regression Machine Learning Models in 5g Wireless Networks

“…В статье [5] исследуется устойчивость коэффициентов регрессии к состязательным примерам, подготовленным для «отравления» исходных данных обучения модели МО. Публикация [6] посвящена анализу уязвимости регрессионных моделей многомерных временных рядов к состязательным атакам. Авторы показывают, что исследуемые модели уязвимы к проводимым атакам, что критически важно для безопасности.…”

Toward native explainable and robust AI in 6G networks: Current state, challenges and road ahead

“…The first category assumes that the adversary has full knowledge of the training data, model architecture, and parameters, the latter none, and gray-box attacks assume partial knowledge. Known attacks on time-series are modifications of attacks originally designed for images like the Fast Gradient Sign Method (FGSM) and its iterative version Basic Iterative Method (BIM) [19]. Both generate a crafted input that resembles the original one, but the values of its elements are equal to the sign of the elements of the gradient of the cost function.…”

Adversarial trading

“…Since the seminal work of (Dalvi, Domingos, Mausam, Sanghai, & Verma, 2004), (Lowd & Meek, 2005) and the more recent (Szegedy et al, 2013), adversarial samples is an active field of research in Machine Learning. It has been explored in computer vision (Moosavi-Dezfooli, Fawzi, & Frossard, 2015), human vision (Elsayed et al, 2018), real 3D objects (Athalye, Engstrom, Ilyas, & Kwok, 2017), tabular data (Ballet et al, 2019), time series (Mode & Hoque, 2020), speech (Carlini & Wagner, 2018) or from a theoretical point of view in (Dohmatob, 2019), (Tsipras, Santurkar, Engstrom, Turner, & Madry, 2018), or (Fawzi, Fawzi, & Frossard, 2017). See (Wiyatno et al, 2019) for a review.…”