Block shuffling learning for Deepfake Detection

View on arXiv ← Back to list

Authors: Sitong Liu, Zhichao Lian, Siqi Gu, Liang Xiao

Published: 2022-02-06 17:16:46+00:00

AI Summary

This paper proposes Block Shuffling Learning (BSL), a novel regularization method for deepfake detection that addresses overfitting issues in CNNs. BSL involves intra- and inter-block image shuffling, combined with an adversarial loss to mitigate noise and position restoration to capture semantic associations, resulting in improved generalization and robustness against image transformations.

Abstract

Deepfake detection methods based on convolutional neural networks (CNN) have demonstrated high accuracy. textcolor{black}{However, these methods often suffer from decreased performance when faced with unknown forgery methods and common transformations such as resizing and blurring, resulting in deviations between training and testing domains.} This phenomenon, known as overfitting, poses a significant challenge. To address this issue, we propose a novel block shuffling regularization method. Firstly, our approach involves dividing the images into blocks and applying both intra-block and inter-block shuffling techniques. This process indirectly achieves weight-sharing across different dimensions. Secondly, we introduce an adversarial loss algorithm to mitigate the overfitting problem induced by the shuffling noise. Finally, we restore the spatial layout of the blocks to capture the semantic associations among them. Extensive experiments validate the effectiveness of our proposed method, which surpasses existing approaches in forgery face detection. Notably, our method exhibits excellent generalization capabilities, demonstrating robustness against cross-dataset evaluations and common image transformations. Especially our method can be easily integrated with various CNN models. Source code is available at href{https://github.com/NoWindButRain/BlockShuffleLearning}{Github}.

Key findings

BSL outperforms existing deepfake detection methods, demonstrating superior performance in cross-dataset evaluations and robustness against common image transformations. The method is easily integrated with various CNN models and effectively addresses overfitting in local image regions.

Approach

The proposed method uses a block shuffling regularization technique to improve the generalization of CNNs for deepfake detection. It shuffles image blocks both within (intra-block) and between (inter-block) blocks, using an adversarial loss to filter out noise from the shuffling and position restoration to maintain semantic information.

Datasets

UNKNOWN

Model(s)

Convolutional Neural Networks (CNNs)

Author countries

UNKNOWN

← Previous