Contrastive learning-based multimodal attention networks for false information detection method

• 1、Artificial Intelligence Security Technology Innovation Team, Beijing Information Science and Technology University, Beijing 100192, China

Abstract：Multi-modal false information has intensively emerged in the cyberspace in recent years and its effective detection is difficult. Facing this problem, this paper focuses on proposing a multi-modal false information detection method based on contrastive learning pre-training and attention mechanism, in which contrastive learning is used to detect different modalities. Feature alignment and potential relationship learning are performed among data, and the attention mechanism is used to realize the interaction between different modal features. Model construction is completed through feature fusion, and ultimately accurate detection of false information is achieved. Compared with the current mainstream methods, the proposed model herein achieves better results in detecting multi-modal false information , and can basically achieve more accurate identification and detection of false information.

Keywords： cyberspace security false information detection multimodal feature fusion contrastive learning attention mechanism