SecureSpectra: Safeguarding Digital Identity from Deep Fake Threats via Intelligent Signatures

View on arXiv ← Back to list

Authors: Oguzhan Baser, Kaan Kale, Sandeep P. Chinchali

Published: 2024-07-01 02:36:27+00:00

AI Summary

SecureSpectra embeds irreversible signatures in audio to defend against deepfake threats, leveraging the inability of deepfake models to replicate high-frequency content. Differential privacy protects signatures from reverse engineering, achieving high detection accuracy with minimal performance compromise.

Abstract

Advancements in DeepFake (DF) audio models pose a significant threat to voice authentication systems, leading to unauthorized access and the spread of misinformation. We introduce a defense mechanism, SecureSpectra, addressing DF threats by embedding orthogonal, irreversible signatures within audio. SecureSpectra leverages the inability of DF models to replicate high-frequency content, which we empirically identify across diverse datasets and DF models. Integrating differential privacy into the pipeline protects signatures from reverse engineering and strikes a delicate balance between enhanced security and minimal performance compromises. Our evaluations on Mozilla Common Voice, LibriSpeech, and VoxCeleb datasets showcase SecureSpectra's superior performance, outperforming recent works by up to 71% in detection accuracy. We open-source SecureSpectra to benefit the research community.

Key findings

SecureSpectra outperforms existing methods by up to 71% in detection accuracy. The integration of differential privacy slightly reduces accuracy but enhances security against reverse engineering. Deepfake models struggle to replicate high-frequency audio content.

Approach

SecureSpectra embeds an irreversible signature in the high-frequency components of audio using a U-Net based model. A separate verification model detects the presence of this signature without revealing its details. Differential privacy is integrated to protect the signature key.

Datasets

Mozilla Common Voice, LibriSpeech, VoxCeleb

Model(s)

U-Net (for signature embedding), 7-layer CNN (for verification)

Author countries

USA, Turkey

← Previous