Litcius/Paper detail

On End-to-End White-Box Adversarial Attacks in Music Information Retrieval

Katharina Prinz, Arthur Flexer, Gerhard Widmer

2021Transactions of the International Society for Music Information Retrieval18 citationsDOIOpen Access PDF

Abstract

Small adversarial perturbations of input data can drastically change the performance of machine learning systems, thereby challenging their validity. We compare several adversarial attacks targeting an instrument classifier, where for the first time in Music Information Retrieval (MIR) the perturbations are computed directly on the waveform. The attacks can reduce the accuracy of the classifier significantly, while at the same time keeping perturbations almost imperceptible. Furthermore, we show the potential of adversarial attacks being a security issue in MIR by artificially boosting playcounts through an attack on a real-world music recommender system.

Topics & Concepts

Adversarial systemClassifier (UML)Computer scienceBoosting (machine learning)End-to-end principleArtificial intelligenceAdversarial machine learningMusic information retrievalMachine learningSpeech recognitionMusicalVisual artsArtAdversarial Robustness in Machine LearningAnomaly Detection Techniques and ApplicationsMusic and Audio Processing