Deepfake Detection

• Grant number: 25/15682-7
• Support Opportunities: Scholarships in Brazil - Doctorate (Direct)
• Start date: September 01, 2025
• End date: August 31, 2029
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Anderson de Rezende Rocha
• Grantee: Mateus de Padua Vicente
• Host Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Associated research grant: 23/12865-8 - Horus: artificial intelligence techniques to detect and forestall synthetic realities, AP.TEM

Abstract

This Ph.D. research has two key objectives. We aim to design deepfake detection methods focusing primarily on images and videos through comprehensive and scalable formulations. The focus is on open-set scenarios, in which deepfake detection techniques typically struggle to maintain performance due to unfamiliar domains - unknown methods used for synthesis. The objective is to enhance detection techniques by utilizing autoencoders and fusion techniques to generate robust features for bonafide media, along with considering time-dependent artifacts and interactions between images and subsequent frames in videos.The second objective is to incorporate contextual information such as audio. Deepfake videos often retain their original audio, which may be inconsistent with the altered image (e.g., lips andfacial expressions portraying contrasting emotions to the audio). On this front, our goal is to design a multi-modality input deepfake detection method capable of dealing with images, videos, and audio information before decision-making.