ArtifactNet: Detecting AI-Generated Music via Forensic Residual Physics
2026-04-17 • Sound
Sound
AI summaryⓘ
The authors introduce ArtifactNet, a small and efficient system that can tell if music was made by AI by looking for hidden technical clues left by audio compression tools. Their method uses a special neural network to find leftover signals in the sound waves and then classifies these clues with another lightweight model. They also created a testing dataset called ArtifactBench with many AI and real music samples to fairly check how well their system works. ArtifactNet performs much better than previous methods, especially when music is compressed in different ways, making it a reliable and lightweight way to detect AI-generated music.
AI-generated musicneural audio codecsUNetmagnitude spectrogramHPSS (Harmonic-Percussive Source Separation)CNN (Convolutional Neural Network)zero-shot evaluationfalse positive ratecross-codec augmentationforensic physics
Authors
Heewon Oh
Abstract
We present ArtifactNet, a lightweight framework that detects AI-generated music by reframing the problem as forensic physics -- extracting and analyzing the physical artifacts that neural audio codecs inevitably imprint on generated audio. A bounded-mask UNet (ArtifactUNet, 3.6M parameters) extracts codec residuals from magnitude spectrograms, which are then decomposed via HPSS into 7-channel forensic features for classification by a compact CNN (0.4M parameters; 4.0M total). We introduce ArtifactBench, a multi-generator evaluation benchmark comprising 6,183 tracks (4,383 AI from 22 generators and 1,800 real from 6 diverse sources). Each track is tagged with bench_origin for fair zero-shot evaluation. On the unseen test partition (n=2,263), ArtifactNet achieves F1 = 0.9829 with FPR = 1.49%, compared to CLAM (F1 = 0.7576, FPR = 69.26%) and SpecTTTra (F1 = 0.7713, FPR = 19.43%) evaluated under identical conditions with published checkpoints. Codec-aware training (4-way WAV/MP3/AAC/Opus augmentation) further reduces cross-codec probability drift by 83% (Delta = 0.95 -> 0.16), resolving the primary codec-invariance failure mode. These results establish forensic physics -- direct extraction of codec-level artifacts -- as a more generalizable and parameter-efficient paradigm for AI music detection than representation learning, using 49x fewer parameters than CLAM and 4.8x fewer than SpecTTTra.