Bridging the Age Gap: Towards Detecting Neural Audio Codec Synthesized Elderly Speech Deepfake

2026-06-19

AI Summary

This study addresses the limited generalization of existing deepfake detection methods on elderly speech, particularly against forgeries synthesized by neural audio codecs. To tackle this challenge, the authors introduce a new task termed Elderly Codec Forgery Detection (ECFD), construct the first bilingual (Chinese–English) ECF dataset, and propose BONSAI—a novel framework leveraging multimodal foundation models LanguageBind and ImageBind. BONSAI fuses cross-modal features via Jensen–Shannon divergence, achieving an average Equal Error Rate (EER) of 1.66% on the ECFD task. This performance significantly surpasses both unimodal models and current state-of-the-art baselines, establishing a new benchmark for detecting voice forgeries targeting elderly speakers.

Abstract

In this study, we introduce the Elderly CodecFake Detection (ECFD) task and release the Elderly-CodecFake (ECF) dataset in English and Chinese. We show that state-of-the-art CF detectors trained on previous benchmark CF datasets generalize poorly to elderly speech, revealing a critical vulnerability. We further hypothesize and demonstrate that multimodal foundation models (FMs) such as LanguageBind (LB) and ImageBind (IB) are more effective for ECFD due to their exposure to elderly content during cross-modal pretraining. Motivated by prior evidence that fusion of FMs enhances downstream performance, we explore fusion of FMs for ECFD. To this end, we propose BONSAI, a novel framework that employs Jensen-Shannon Divergence as the fusion mechanism. BONSAI with the fusion of LB and IB achieves an average EER (%) of 1.66 and outperforms individual FMs as well as competitive SOTA baselines, establishing a new benchmark for the ECFD task.

Problem

• elderly speech
• deepfake detection
• audio codec
• generalization gap
• synthetic speech

Innovation

• elderly deepfake detection
• neural audio codec
• multimodal foundation models
• model fusion
• Jensen-Shannon Divergence