Deepfake Representation with Multilinear Regression

View on arXiv ← Back to list

Authors: Sara Abdali, M. Alex O. Vasilescu, Evangelos E. Papalexakis

Published: 2021-08-15 09:37:38+00:00

AI Summary

This paper proposes a modified multilinear (tensor) method, combining linear and multilinear regressions, for representing deepfake and real data. The approach uses this representation to perform SVM classification of deepfakes, achieving encouraging results.

Abstract

Generative neural network architectures such as GANs, may be used to generate synthetic instances to compensate for the lack of real data. However, they may be employed to create media that may cause social, political or economical upheaval. One emerging media is Deepfake.Techniques that can discriminate between such media is indispensable. In this paper, we propose a modified multilinear (tensor) method, a combination of linear and multilinear regressions for representing fake and real data. We test our approach by representing Deepfakes with our modified multilinear (tensor) approach and perform SVM classification with encouraging results.

Key findings

The proposed multilinear regression approach shows encouraging results in deepfake classification. The focus on outer facial rings, where artifacts are concentrated, improves the effectiveness of the model. Further details on the quantitative results are unavailable in the provided abstract and introduction.

Approach

The authors segment facial images into inner and outer rings, focusing on the outer ring where deepfake artifacts are concentrated. They use a modified multilinear tensor model to represent these regions for real and fake data, followed by SVM classification.

Datasets

UNKNOWN

Model(s)

SVM classifier using a modified multilinear (tensor) representation of facial image regions.

Author countries

USA

← Previous