MAIA fools music AI systems by locating the most decision-critical fragments of a track and regenerating them with a generative inpainting model, in both white-box and black-box settings.
This framework reconstructs critical audio segments with adversarial perturbations, ensuring musical coher- ence while effectively misleading target models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SD 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MAIA: An Inpainting-Based Approach for Music Adversarial Attacks
MAIA fools music AI systems by locating the most decision-critical fragments of a track and regenerating them with a generative inpainting model, in both white-box and black-box settings.