Universal Fourier Attack for Time Series
This addresses the problem of implementing adversarial attacks in real-world scenarios for time series applications, representing an incremental advancement by adapting existing attack concepts to new constraints.
The paper tackles the challenge of creating real-world adversarial attacks for time series data by introducing a universal, time-invariant attack with a frequency spectrum matching the original data, demonstrating effectiveness in speech recognition and unintended radiated emission domains with robustness against common defenses.
A wide variety of adversarial attacks have been proposed and explored using image and audio data. These attacks are notoriously easy to generate digitally when the attacker can directly manipulate the input to a model, but are much more difficult to implement in the real-world. In this paper we present a universal, time invariant attack for general time series data such that the attack has a frequency spectrum primarily composed of the frequencies present in the original data. The universality of the attack makes it fast and easy to implement as no computation is required to add it to an input, while time invariance is useful for real-world deployment. Additionally, the frequency constraint ensures the attack can withstand filtering. We demonstrate the effectiveness of the attack in two different domains, speech recognition and unintended radiated emission, and show that the attack is robust against common transform-and-compare defense pipelines.