Teffic-Audio system advances speech deepfake detection accuracy

Researchers have developed Teffic-Audio, a new system designed to detect sophisticated speech deepfakes. The system utilizes a Conformer-based encoder and a binary classifier, improving generalization through its training methodology which includes multi-source data, balanced sampling, and diverse audio augmentation. Teffic-Audio achieved a 1.454% EER on the Speech-DF-Arena benchmark, outperforming existing public systems and demonstrating a favorable performance-complexity trade-off. AI

IMPACT Improves the accuracy and generalization of speech deepfake detection systems.

RANK_REASON Research paper detailing a new system for speech deepfake detection. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• Conformer
• DagsHub
• Gotit.pub
• Hugging Face
• ScienceCast
• Speech-DF-Arena
• Teffic-Audio
• Wan Lin

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →