AudioStealer: Extracting Audio Prompts via Shapley Value-Guided Query Search

As text-to-music models gain widespread adoption, the prompts used to guide these systems have become valuable intellectual property. This shift has given rise to a new form of attack: prompt stealing, aiming to reconstruct the high-value prompts that guide music generation. However, unlike prior work in text and image generation, prompt stealing in textto-music systems faces unique challenges due to the entangled and diffuse nature of semantic representations in audio, which complicates decoupling textual tokens from acoustic outputs. To address these challenges, we present AudioStealer, the first study of prompt inversion in the audio domain. AudioStealer operates via a two-stage black-box attack framework: first, a heuristic search guided by audiolanguage embeddings identifies initial candidates; then, these candidates are refined using a game-theoretic strategy based on Shapley value estimation to attribute semantic contributions. Our method requires no direct access to the target model and relies solely on a shadow model, making it applicable. Through extensive experiments, we demonstrate that AudioStealer recovers prompts with high textual consistency to the ground truth, while regenerated audio maintains strong perceptual similarity to target recordings. These results expose vulnerabilities in the text-to-audio market ecosystem and underscore the need for intellectual property protections in generative audio technologies. Our code and dataset are openly available at https: //github.com/kprisoner/AudioStealer .