Preliminary Forensics Analysis of DeepFake Images

View on arXiv ← Back to list

Authors: Luca Guarnera, Oliver Giudice, Cristina Nastasi, Sebastiano Battiato

Published: 2020-04-27 08:09:06+00:00

AI Summary

This paper explores the detection of deepfake images by analyzing anomalies in the frequency domain. It finds that standard image forensics techniques alone are insufficient, but analyzing Fourier transforms reveals patterns specific to different deepfake generation technologies.

Abstract

One of the most terrifying phenomenon nowadays is the DeepFake: the possibility to automatically replace a person's face in images and videos by exploiting algorithms based on deep learning. This paper will present a brief overview of technologies able to produce DeepFake images of faces. A forensics analysis of those images with standard methods will be presented: not surprisingly state of the art techniques are not completely able to detect the fakeness. To solve this, a preliminary idea on how to fight DeepFake images of faces will be presented by analysing anomalies in the frequency domain.

Key findings

Standard image forensics tools show some anomalies in deepfake images but are not sufficient for definitive detection. Analysis of Fourier transforms reveals characteristic frequency patterns for different deepfake generation techniques. These findings suggest potential for improved deepfake detection.

Approach

The authors propose analyzing deepfake images in the frequency domain after applying Fourier transforms. They observe characteristic patterns in the frequency domain associated with specific deepfake generation methods, suggesting a new avenue for detection.

Datasets

Images generated by StyleGAN and StarGAN.

Model(s)

UNKNOWN. The paper does not specify models used for detection, but rather focuses on a novel analysis technique applied to existing deepfake images.

Author countries

Italy

← Previous