SONICS: Synthetic Or Not -- Identifying Counterfeit Songs

TL;DR

This paper introduces SONICS, a large dataset for detecting AI-generated synthetic songs, emphasizing the importance of modeling long-range dependencies with a new efficient architecture, SpecTTTra, to improve detection accuracy and efficiency.

Contribution

The paper presents SONICS, a comprehensive dataset for end-to-end fake song detection, and proposes SpecTTTra, an innovative architecture that enhances long-range temporal modeling with better efficiency.

Findings

SpecTTTra outperforms ViT by 8% in F1 score on long songs.

SpecTTTra is 38% faster and uses 26% less memory than CNN and Transformer models.

The SONICS dataset includes over 97k songs, addressing previous dataset limitations.

Abstract

The recent surge in AI-generated songs presents exciting possibilities and challenges. These innovations necessitate the ability to distinguish between human-composed and synthetic songs to safeguard artistic integrity and protect human musical artistry. Existing research and datasets in fake song detection only focus on singing voice deepfake detection (SVDD), where the vocals are AI-generated but the instrumental music is sourced from real songs. However, these approaches are inadequate for detecting contemporary end-to-end artificial songs where all components (vocals, music, lyrics, and style) could be AI-generated. Additionally, existing datasets lack music-lyrics diversity, long-duration songs, and open-access fake songs. To address these gaps, we introduce SONICS, a novel dataset for end-to-end Synthetic Song Detection (SSD), comprising over 97k songs (4,751 hours) with over 49k synthetic songs from popular platforms like Suno and Udio. Furthermore, we highlight the importance of modeling long-range temporal dependencies in songs for effective authenticity detection, an aspect entirely overlooked in existing methods. To utilize long-range patterns, we introduce SpecTTTra, a novel architecture that significantly improves time and memory efficiency over conventional CNN and Transformer-based models. For long songs, our top-performing variant outperforms ViT by 8% in F1 score, is 38% faster, and uses 26% less memory, while also surpassing ConvNeXt with a 1% F1 score gain, 20% speed boost, and 67% memory reduction.