Towards Unconstrained Audio Splicing Detection and Localization with Neural Networks

TL;DR

This paper introduces a Transformer-based neural network approach for detecting and localizing audio splicing in unconstrained, real-world scenarios, outperforming existing methods and addressing the limitations of handcrafted feature-based algorithms.

Contribution

It presents a novel Transformer seq2seq model for audio splicing detection and localization, capable of handling unconstrained audio samples with various post-processing disguises.

Findings

Outperforms existing dedicated splicing detection methods

Superior to general-purpose networks like EfficientNet and RegNet

Effective in simulated attack scenarios with post-processing operations

Abstract

Freely available and easy-to-use audio editing tools make it straightforward to perform audio splicing. Convincing forgeries can be created by combining various speech samples from the same person. Detection of such splices is important both in the public sector when considering misinformation, and in a legal context to verify the integrity of evidence. Unfortunately, most existing detection algorithms for audio splicing use handcrafted features and make specific assumptions. However, criminal investigators are often faced with audio samples from unconstrained sources with unknown characteristics, which raises the need for more generally applicable methods. With this work, we aim to take a first step towards unconstrained audio splicing detection to address this need. We simulate various attack scenarios in the form of post-processing operations that may disguise splicing. We propose a Transformer sequence-to-sequence (seq2seq) network for splicing detection and localization. Our extensive evaluation shows that the proposed method outperforms existing dedicated approaches for splicing detection [3, 10] as well as the general-purpose networks EfficientNet [28] and RegNet [25].