Towards Open-world Generalized Deepfake Detection: General Feature Extraction via Unsupervised Domain Adaptation

View on arXiv ← Back to list

Authors: Midou Guo, Qilin Yin, Wei Lu, Xiangyang Luo

Published: 2025-05-18 10:12:12+00:00

AI Summary

This paper introduces a novel Open-World Deepfake Detection Generalization Enhancement Training Strategy (OWG-DS) to improve deepfake detection in open-world scenarios with limited labeled data. OWG-DS leverages unsupervised domain adaptation, aligning features from labeled and unlabeled data through Domain Distance Optimization and Similarity-based Class Boundary Separation modules, and utilizes adversarial training for robust feature learning.

Abstract

With the development of generative artificial intelligence, new forgery methods are rapidly emerging. Social platforms are flooded with vast amounts of unlabeled synthetic data and authentic data, making it increasingly challenging to distinguish real from fake. Due to the lack of labels, existing supervised detection methods struggle to effectively address the detection of unknown deepfake methods. Moreover, in open world scenarios, the amount of unlabeled data greatly exceeds that of labeled data. Therefore, we define a new deepfake detection generalization task which focuses on how to achieve efficient detection of large amounts of unlabeled data based on limited labeled data to simulate a open world scenario. To solve the above mentioned task, we propose a novel Open-World Deepfake Detection Generalization Enhancement Training Strategy (OWG-DS) to improve the generalization ability of existing methods. Our approach aims to transfer deepfake detection knowledge from a small amount of labeled source domain data to large-scale unlabeled target domain data. Specifically, we introduce the Domain Distance Optimization (DDO) module to align different domain features by optimizing both inter-domain and intra-domain distances. Additionally, the Similarity-based Class Boundary Separation (SCBS) module is used to enhance the aggregation of similar samples to ensure clearer class boundaries, while an adversarial training mechanism is adopted to learn the domain-invariant features. Extensive experiments show that the proposed deepfake detection generalization enhancement training strategy excels in cross-method and cross-dataset scenarios, improving the model's generalization.

Key findings

OWG-DS significantly outperforms existing methods in cross-method and cross-dataset deepfake detection scenarios. It demonstrates superior generalization ability, particularly with limited labeled data in the target domain. The ablation study confirms the importance of each module in the proposed strategy.

Approach

The proposed OWG-DS uses a three-module approach: Domain Distance Optimization (DDO) to align features across domains, Similarity-based Class Boundary Separation (SCBS) to refine class boundaries, and adversarial training to learn domain-invariant features. This strategy transfers knowledge from a small labeled dataset to a large unlabeled dataset, improving generalization.

Datasets

FF++ (FaceForensics++), Celeb-DF, DFDC

Model(s)

Xception, ResNet-50, EfficientNet-B0 (primarily Xception in the main experiments)

Author countries

China

← Previous