Fighting deepfakes by detecting GAN DCT anomalies

View on arXiv ← Back to list

Authors: Oliver Giudice, Luca Guarnera, Sebastiano Battiato

Published: 2021-01-24 19:45:11+00:00

AI Summary

This paper proposes a novel deepfake detection method, CTF-DCT, that analyzes the Discrete Cosine Transform (DCT) coefficients of images to identify GAN-specific frequencies (GSFs), acting as a unique fingerprint for different generative architectures. The method uses beta statistics from the AC coefficient distribution to distinguish real from fake images, exceeding state-of-the-art methods in accuracy and explainability.

Abstract

To properly contrast the Deepfake phenomenon the need to design new Deepfake detection algorithms arises; the misuse of this formidable A.I. technology brings serious consequences in the private life of every involved person. State-of-the-art proliferates with solutions using deep neural networks to detect a fake multimedia content but unfortunately these algorithms appear to be neither generalizable nor explainable. However, traces left by Generative Adversarial Network (GAN) engines during the creation of the Deepfakes can be detected by analyzing ad-hoc frequencies. For this reason, in this paper we propose a new pipeline able to detect the so-called GAN Specific Frequencies (GSF) representing a unique fingerprint of the different generative architectures. By employing Discrete Cosine Transform (DCT), anomalous frequencies were detected. The BETA statistics inferred by the AC coefficients distribution have been the key to recognize GAN-engine generated data. Robustness tests were also carried out in order to demonstrate the effectiveness of the technique using different attacks on images such as JPEG Compression, mirroring, rotation, scaling, addition of random sized rectangles. Experiments demonstrated that the method is innovative, exceeds the state of the art and also give many insights in terms of explainability.

Key findings

The CTF-DCT method demonstrates superior performance compared to existing techniques, achieving high accuracy in distinguishing real from deepfake images. The approach provides explainability by identifying specific frequencies associated with GAN architectures. Robustness tests show the method's resilience to common image manipulations.

Approach

The CTF-DCT method applies a DCT to 8x8 blocks of an image. It then models the AC coefficients' distribution with a Laplacian distribution and extracts beta statistics as features. A classifier, not explicitly specified but implied to be a gradient boosting model, uses these features to distinguish between real and deepfake images.

Datasets

CelebA, FFHQ, datasets generated using StarGAN, AttGAN, GDWCT, StyleGAN, and StyleGAN2 architectures.

Model(s)

Gradient boosting model (implied); DCT is used for feature extraction, not as a model itself.

Author countries

Italy

← Previous