Face Swap Detection: A Systematic Literature Review
Safaa Hriez · IEEE Access · 2025
Face swap technology, often associated with deepfakes, has rapidly advanced in recent years, raising serious concerns around misinformation, digital impersonation, and privacy. As a result, the development of reliable face swap detection methods has become a critical area of research. This survey provides a comprehensive review of existing approaches to face swap detection, addressing key research questions such as commonly used datasets, evaluation metrics, comparative model performance, and persistent challenges in the field. It includes a detailed taxonomy of detection methods, categorizing approaches into spatial, temporal, and spatiotemporal techniques. The survey further examines cross-dataset generalization performance to assess how well models adapt to domain shifts between training and testing data. Recent innovative directions are explored, covering adversarial defense strategies, explainability techniques, lightweight models for edge deployment, and privacy-preserving training. Additionally, best practices for building and releasing face swap detection tools are discussed to promote ethical, robust, and practical implementations. Finally, the paper outlines future research directions aimed at advancing model robustness, generalization, and compliance with legal and ethical standards. The discussions provide valuable insights that help researchers and practitioners gain a clear understanding of the face swap detection field, supporting and guiding their future research efforts.