Search papers, labs, and topics across Lattice.
This paper tackles the vulnerability of AI-generated music detectors to simple audio manipulations like speed modification and pitch shifting, which previously led to significant performance drops. By introducing a frequency-scaling-invariant detection pipeline that employs log-STFT remapping and a learned cross-correlation filter, the authors enhance the robustness of detection systems against these manipulations. The proposed method not only achieves high accuracy in identifying AI-generated music but also provides interpretable outputs, including an estimate of the speed-change factor applied to the audio.
AI music detectors can now maintain high accuracy even when faced with common audio manipulations, thanks to a novel frequency-scaling-invariant approach.
AI music generators leave predictable spectral artifacts determined by their architecture. Existing detectors exploit these artifacts with near-perfect accuracy on raw generated tracks, but their performance collapses under simple audio manipulations, such as speed modification or pitch shifting. We address this open robustness problem by introducing a frequency-scaling-invariant detection pipeline that aims to prevent this kind of attack by design. Our method maps audio onto a log-frequency axis via a log-STFT remapping. A single learned cross-correlation filter, combined with max-pooling, provides shift invariance at inference time. Training uses a hybrid loss that jointly supervises binary detection and artifact-peak localization, regularizing boundary weights. Because robustness to speed change is built in by design, the detector is also interpretable: it outputs both a binary decision and an estimate of the applied speed-change factor.
