Koreanische Polizei, Wuppertal

Authenticity System The widespread availability of generative AI tools has lowered the barrier to create highly realistic manipulated content, such as deepfake videos and synthetic voices.
This poses a risk of misuse, including fraud, misinformation, and other cybercrimes.
Current detection methods are struggling to keep up with the fast development of generative AI, and detection processes remain heavily reliant on manual review, making them inefficient and unsustainable. Furthermore, existing deepfake detection models often suffer from biased or insufficient datasets, leading to performance gaps when applied to real-world cases. To address these challenges, this project aims to develop a state-of-the-art deepfake dataset focused on AI-generated manipulations. This dataset will enhance model training, evaluation, and improve the robustness of detection systems against emerging threats.
The primary goal of this project is to develop a state-of-the-art multimodal deepfake dataset to improve the detection of AI-generated manipulations, with a special focus on Western data sources. The dataset will include both single-modality fakes (i.e., image-only, text-only, speech-only) and synchronized multimodal fakes (i.e., image+text, video+voice). It will cover diverse content from different domains (e.g., politics, business, social media) and feature real people with different