Early detection of fake news on virtual social networks: A time-aware approach based on crowd signals

Abstract

With virtual social networks (VSNs) ever more popular as a means of disseminating fake news, methods for automatically detecting this type of content have become increasingly important. Early fake news detection seeks to detect a fake news story before it is widely spread to enable the implementation of mitigatory actions to minimize its adverse effects on society. Studies based on early detection are promising, but use restricted-access data (e.g., personal and sensitive data) from users’ profiles, which is typically not available in VSNs. To address the problem of restricted data, albeit leaving aside early detection, some studies have explored using crowd signals. This approach combines the opinions (i.e., signals) of a group of users (i.e., the crowd) to detect fake news. Given this scenario, this paper hypothesizes that using users’ opinions can enable the early detection of fake news with results comparable to those of existing state-of-the-art methods, but without relying on restricted-access data. Thus, we propose a novel fake news early detection method: TCS (Time-aware Crowd Signals). The proposed method explores the temporal nature of news propagation to detect fake news, utilizing users’ reputations obtained from their public behavior when spreading news in the past. Experiments on four different datasets show that the proposed method performance is comparable with state-of-the-art methods. As well as confirming the hypothesis, TCS identified fake news stories in VSNs earlier than other methods without relying on restricted-access data from network users.