DisinfoNet: A Comprehensive Dataset of Disinformation with Contents and Propagation Dynamics Over Social Networks

With the continuous development of the Internet, social networks have become crucial platforms for public communication. In this context, disinformation has increasingly proliferated on social media platforms. To address this challenge, two primary technical approaches can be employed: disinformation detection techniques to verify content authenticity, and disinformation source localization methods to trace propagation origins. However, existing datasets face significant limitations, They often lack authenticity in construction methodology and suffer from insufficient scale to accurately represent real-world social network dynamics. In this paper, we introduce a novel dataset, DisinfoNet, constructed by first retrieving claims that have been verified as false by reputable fact-checking websites. We then retain only those claims for which the original posts are still accessible on Twitter, yielding a final corpus of 435 distinct pieces of disinformation. Subsequently, we collected their complete propagation networks to capture the underlying dissemination patterns. We provide a comprehensive characterization of the DisinfoNet dataset and demonstrate its utility through multifaceted experimental studies. Furthermore, we discuss potential applications and future research directions enabled by this dataset.