Robust audio deepfake detection improves with adversarial workflow design
Robust Workflow Generation via Adversarial Learning for Audio Deepfake Detection
SoundArtificial Intelligence
Summary
Audio deepfakes are fake voice clips that sound real and can be used for fraud or trickery. Current tools can detect these fakes in ideal settings but struggle when the audio is noisy or altered in real life. The authors propose a system called ROGUE that learns to pick the best combination of detection tools while also simulating tough conditions to train on. This makes detecting fake voices more reliable, even when the audio is degraded or manipulated. Tests show ROGUE works better than previous methods on various datasets and noisy audio.
What this means in practice
- •For cybersecurity teams: Deploy workflows that adaptively combine detection methods to identify audio deepfake threats under varied real-world noise and distortions.
- •For voice assistant developers: Integrate adaptive detection workflows into voice-controlled devices to improve security against manipulated synthetic speech commands.
Authors
Xiang Li, Pin-Yu Chen, Wenqi Wei
Abstract
The rapid advancement of speech synthesis and voice conversion technologies has made audio deepfakes increasingly realistic, posing serious security risks in practical applications. While existing detection methods achieve strong performance under controlled conditions, they often fail to generalize under real-world perturbations and corruptions. In this paper, we propose ROGUE, a framework that dynamically constructs robust detection workflows by orchestrating multiple detection tools. ROGUE formulates workflow generation as a sequential decision-making problem and introduces a dual-agent paradigm, where a perturbation agent generates audio perturbations and a policy agent learns to select and execute detection tools under perturbed conditions. Through adversarial learning, ROGUE enables perturbation-aware tool selection, adaptive execution strategies, and improved robustness to distribution shifts. Extensive experiments across multiple datasets and real-world corruptions demonstrate that ROGUE consistently outperforms strong baselines in both robustness and generalization. Our results highlight the effectiveness of adversarially optimized workflow generation for building reliable audio deepfake detection systems in real-world deployment settings.