Does Audio Deepfake Detection Generalize?
This work addresses the generalization issue in audio deepfake detection for security applications, revealing that current methods are overly tailored to benchmarks and perform poorly in real-world scenarios, which is an incremental but important finding.
The paper tackled the problem of audio deepfake detection by systematically re-implementing and evaluating existing architectures to identify key factors for success, finding that using cqtspec or logspec features improves performance by 37% EER on average, but performance degrades by up to 1000% on real-world data.
Current text-to-speech algorithms produce realistic fakes of human voices, making deepfake detection a much-needed area of research. While researchers have presented various techniques for detecting audio spoofs, it is often unclear exactly why these architectures are successful: Preprocessing steps, hyperparameter settings, and the degree of fine-tuning are not consistent across related work. Which factors contribute to success, and which are accidental? In this work, we address this problem: We systematize audio spoofing detection by re-implementing and uniformly evaluating architectures from related work. We identify overarching features for successful audio deepfake detection, such as using cqtspec or logspec features instead of melspec features, which improves performance by 37% EER on average, all other factors constant. Additionally, we evaluate generalization capabilities: We collect and publish a new dataset consisting of 37.9 hours of found audio recordings of celebrities and politicians, of which 17.2 hours are deepfakes. We find that related work performs poorly on such real-world data (performance degradation of up to one thousand percent). This may suggest that the community has tailored its solutions too closely to the prevailing ASVSpoof benchmark and that deepfakes are much harder to detect outside the lab than previously thought.