Evaluating the Effectiveness of Deepfake Detection Techniques

As deep learning technologies continue to evolve, the rise of deepfake videos has become a critical concern, posing threats to cybersecurity, privacy, and the spread of misinformation. Detecting deepfake content has thus emerged as a crucial research area, leading to the development of several detection techniques. However, the effectiveness and robustness of these techniques vary widely, necessitating a comprehensive evaluation to identify strengths, weaknesses, and potential improvements.

The aim of this research is to evaluate the effectiveness of various deepfake detection techniques through a comparative analysis and the development of a methodological framework. The study will involve gathering different datasets containing both authentic and deepfake videos across various domains and applications. State-of-the-art deepfake detection methods, including traditional approaches (e.g., image forensics, facial recognition, etc.) and deep learning-based models (e.g., convolutional neural networks, generative adversarial networks, etc.), will then be implemented and evaluated based on performance metrics such as accuracy, precision, recall, and F1 score.

Also, the study will propose a methodological framework to help researchers select, benchmark, and validate deepfake detection techniques. The student is expected to propose a methodological framework for deepfake detection and deliver a project report that should address the following things:

• Collect diverse datasets containing authentic and deepfake videos across different domains and applications.
• Implement and evaluate state-of-the-art deepfake detection techniques, including traditional methods and deep learning-based models.
• Compare the performance of different detection techniques based on metrics such as accuracy, precision, recall, and F1 score.
• Develop a methodological framework for evaluating and benchmarking deepfake detection techniques, considering factors such as dataset selection, evaluation metrics, and model validation techniques. 

How to apply

Formal applications can be submitted via the University of Bradford web site; applicants will need to register an account and select ‘Full-time PhD in Computer Science’ as the course, and then specify the project title in the ‘Research Proposal’ section.

About the University of Bradford

Bradford is a research-active University supporting the highest-quality research. We excel in applying our research to benefit our stakeholders by working with employers and organisations world-wide across the private, public, voluntary and community sectors and actively encourage and support our postgraduate researchers to engage in research and business development activities.

Positive Action Statement

At the University of Bradford our vision is a world of inclusion and equality of opportunity, where people want to, and can, make a difference. We place equality and diversity, inclusion, and a commitment to social mobility at the centre of our mission and ethos. In working to make a difference we are committed to addressing systemic inequality and disadvantages experienced by Black, Asian and Minority Ethnic staff and students.

Under sections 158-159 of the Equality Act 2010, positive action can be taken where protected group members are under-represented. At Bradford, our data show that people from Black, Asian, and Minority Ethnic groups who are UK nationals are significantly under-represented at the postgraduate researcher level. 

These are lawful measures designed to address systemic and structural issues which result in the under-representation of Black, Asian, and Minority Ethnic students in PGR studies.