Forensic Challenges in Face Manipulated Videos

Rapid advancements in deepfake generation techniques have significantly enabled highly realistic video manipulations, including face swapping and lip synchronization, while simultaneously raising serious ethical and security concerns. The detection of such deepfake videos has emerged as a critical challenge in digital forensics. However, an equally important but less explored aspect is the identification of generative model. In this paper, findings related to the attribution of face-manipulated videos are presented, where distinguishing artifacts introduced by different deepfake generation methods are analyzed and exploited for forensic investigations. The attribution of a deepfake video to a specific source is crucial for investigations, misinformation mitigation, and forensic analysis. Identifying the source of deep-fakes plays a crucial role in countering their malicious use, and the results contribute to strengthening the reliability of digital forensics in an age increasingly shaped by AI-generated content.