This program aims to cultivate next-generation global talent in AI-driven cybersecurity and misinformation defense by integrating advanced research, professional training, and cross-cultural collaboration. The initiative responds to the global rise of AI-generated misinformation and deepfake media, which threaten social stability and digital trust across online ecosystems and metaverse platforms. Through a combination of hands-on learning, interdisciplinary research, and cultural immersion, participants will gain practical expertise in data collection, multimodal model design, adversarial robustness testing, and ethical AI evaluation—advancing Taiwan's role as a regional innovation hub for trustworthy AI technologies. Technical Framework The project comprises three integrated layers that collectively ensure end-to-end detection and validation of misinformation: 1. Textual Threat Analysis Layer: Utilizes fine-tuned transformer models (DeBERTa-v3, LLaMA-2, GPT-style encoders) to detect linguistic inconsistencies, semantic drift, and emotionally biased manipulation in multilingual text corpora. 2. Multimodal Fusion Layer: Integrates CNN-LSTM and Vision Transformer (ViT) architectures for cross-modal verification of fake videos and altered audio streams using temporal-spatial attention and contrastive embeddings. 3. Trust and Provenance Validation Layer: Employs graph neural networks (GNNs) and blockchain-based provenance tracking to authenticate information sources and mitigate coordinated manipulation attacks. Objectives -Develop a deployable AI framework for the detection of textual and visual misinformation. -Train international participants in AI security, deepfake forensics, and ethical data governance. -Strengthen Taiwan's global academic visibility in trustworthy and secure AI research. -Establish sustainable international research and student exchange networks.