DeFakePro: Decentralized DeepFake Attacks Detection using ENF Authentication

View on arXiv ← Back to list

Authors: Deeraj Nagothu, Ronghua Xu, Yu Chen, Erik Blasch, Alexander Aved

Published: 2022-07-22 01:22:11+00:00

AI Summary

DeFakePro is a decentralized deepfake detection system for online video conferencing, using Electrical Network Frequency (ENF) as an environmental fingerprint. It leverages a Proof-of-ENF (PoENF) consensus algorithm to authenticate audio and video streams by comparing their ENF signatures to a ground truth.

Abstract

Advancements in generative models, like Deepfake allows users to imitate a targeted person and manipulate online interactions. It has been recognized that disinformation may cause disturbance in society and ruin the foundation of trust. This article presents DeFakePro, a decentralized consensus mechanism-based Deepfake detection technique in online video conferencing tools. Leveraging Electrical Network Frequency (ENF), an environmental fingerprint embedded in digital media recording, affords a consensus mechanism design called Proof-of-ENF (PoENF) algorithm. The similarity in ENF signal fluctuations is utilized in the PoENF algorithm to authenticate the media broadcasted in conferencing tools. By utilizing the video conferencing setup with malicious participants to broadcast deep fake video recordings to other participants, the DeFakePro system verifies the authenticity of the incoming media in both audio and video channels.

Key findings

DeFakePro achieves similar accuracy to state-of-the-art techniques but with faster processing, suitable for real-time applications. The system effectively detects and localizes deepfakes in both audio and video streams by analyzing ENF discrepancies. The PoENF consensus mechanism demonstrates robustness against Byzantine nodes.

Approach

DeFakePro uses the Electrical Network Frequency (ENF) embedded in audio and video recordings as a fingerprint for authentication. A decentralized Proof-of-ENF (PoENF) consensus algorithm compares the ENF of incoming streams to a ground truth, establishing authenticity and localizing forgeries.

Datasets

UNKNOWN. The paper mentions using DeepFaceLive for generating video deepfakes and Descript for audio deepfakes, but doesn't specify a named dataset.

Model(s)

The paper does not explicitly mention using deep learning models for deepfake detection. The core approach relies on comparing ENF signals.

Author countries

USA

← Previous