Existing audio adversarial attack methods suffer from poor transferability, primarily due to insufficient exploration of model decision mechanisms and overreliance on heuristic-driven algorithm design. This paper aims to alleviate this gap. Specifically, through observations across three mainstream audio tasks (Automatic Speech Recognition, Speaker Verification, and Keyword Spotting), we reveal that these models primarily rely on local temporal features—inputs with time shuffled retain 83.7% of original accuracy. The SHAP-based visualization further validated that time shuffle leads to a significant shift in the salient regions of the model, but the samples can still be correctly identified, indicating the presence of redundant features that can affect decision-making. Inspired by these findings, we propose Time-Shuffle (TS) adversarial attack (including segments-based TS and phoneme-level-based TS-p). This method divides audio or phonemes into segments, randomly shuffles them, and computes gradients on the shuffled structure. By forcing perturbations to exploit transferable local temporal features and reduce overfitting to source-specific patterns, TS/TS-p inherently enhances transferability. As a model-agnostic framework, TS/TS-p can seamlessly integrate with existing attack methods. Comprehensive experiments demonstrate that TS-p achieved SOTA and boosts transferability by about 23%/14.7%/6.3% on ASR/ASV/KWS.

Music deepfake detection is aimed at identifying whether songs are generated by AI. Current methods usually separate vocals from background music for detection, but this could leave residual forgery information in the background. Our study demonstrates for the first time that incorporating background forgery information with vocals can improve detection accuracy. Furthermore, our findings show that using background features can reduce EER by an average of about 2% on existing frameworks. Based on this observation, we propose a novel Hybrid Frontend that captures generalized features from both vocal and background music. The Hybrid Frontend comprises two branches: vocal and background music part. Specifically, the vocal part uses the Sinconv encoder as a deeply embedded feature extractor. The latter captures background variation by fine-tuning the pre-trained model with adapters. Experimental results demonstrate that our method outperforms the vocal-only detection on WildSVDD dataset, achieving an EER of 8.53%, which is 1.3% lower.