Smooth Perturbations for Time Series Adversarial Attacks

Abstract

Adversarial attacks represent a threat to every deep neural network. They are particularly effective if they can perturb a given model while remaining undetectable. They have been initially introduced for image classifiers, and are well studied for this task. For time series, few attacks have yet been proposed. Most that have are adaptations of attacks previously proposed for image classifiers. Although these attacks are effective, they generate perturbations containing clearly discernible patterns such as sawtooth and spikes. Adversarial patterns are not perceptible on images, but the attacks proposed to date are readily perceptible in the case of time series. In order to generate stealthier adversarial attacks for time series, we propose a new attack that produces smoother perturbations. We find that smooth perturbations are harder to detect by the naked eye. We also show how adversarial training can improve model robustness against this attack, thus making models less vulnerable.

Achievable Rate of Near-Field Communications Based on Physically Consistent Models

“…For computer vision applications, the function M(•) in (2) encloses different manual transformations such as altering object properties (e.g., shape, location, texture), scene editing (e.g., illumination, camera view), or random noise injection [37]. For real-valued data input vectors, augmentation involves scaling, pattern switching, and random perturbation [38]. These augmentation methods are particularly interesting for wireless communication applications because they handle general signal transmission scenarios that are tolerant to variations in the path-loss coefficient, synchronization delays, signal-tonoise ratio, etc.…”