End-to-End Adversarial White Box Attacks on Music Instrument Classification

TL;DR

This paper introduces the first end-to-end adversarial attack method on music instrument classification, directly perturbing audio waveforms to cause misclassification while remaining nearly imperceptible.

Contribution

It presents a novel end-to-end attack approach on audio data, bypassing spectrogram-based methods and demonstrating high effectiveness in fooling classifiers.

Findings

Accuracy drops near random chance levels.

Perturbations are almost imperceptible.

Any instrument can be targeted for misclassification.

Abstract

Small adversarial perturbations of input data are able to drastically change performance of machine learning systems, thereby challenging the validity of such systems. We present the very first end-to-end adversarial attacks on a music instrument classification system allowing to add perturbations directly to audio waveforms instead of spectrograms. Our attacks are able to reduce the accuracy close to a random baseline while at the same time keeping perturbations almost imperceptible and producing misclassifications to any desired instrument.