Human-AI Collaborative Framework for Deepfake Detection: Integrating Perceptual-Cognitive Mechanisms with Deep Learning Models
Applications are invited for a 3.5-year PhD studentship. The studentship will start on 1 October 2025.
Project description
The proliferation of deepfake technology presents significant societal and security challenges, particularly in digital forensics and identity verification systems. While current AI-based detection approaches show promise, they face critical limitations in their ability to generalize to new manipulation techniques and lack interpretability in their decision-making processes. This challenge is particularly significant for courts and digital forensic investigations where trust in visual evidence is paramount.
This PhD project aims to overcome these limitations by developing novel approaches to deepfake detection that combine AI with human perceptual and cognitive processes. The core of this project is to develop novel deep learning architectures for deepfake detection. Comparative studies will be conducted between human and AI performance in deepfake detection across various types of manipulated biometric traits, such as face and voice, identifying complementary strengths and weaknesses. Furthermore, the project will evaluate the interpretability and reliability of the developed solutions through both computational metrics and human factors analysis. By the end of the project, a hybrid detection framework that leverages both human perceptual insights and deep learning capabilities will be built.
Key expected outcomes include high-impact publications in renowned conferences and journals and the development of interpretable AI systems that effectively complement human capabilities. The developed framework will establish new methods for human-AI collaboration in digital forensics, providing foundations for future research in trustworthy AI systems.