A Comprehensive review of fake news detection on social media: feature engineering, feature fusion, and future research directions

Abstract

Social media platforms are mainly used for information sharing, connecting with people, and staying updated about the latest events. However, information present on social media is sometimes incorrect, unverified, or misleading. Such information is often termed as fake news. It is deliberately written to deceive the readers. It has the potential to change their perception about the topic or content being discussed. The best medium to share fake news is social media platforms. Large amounts of misleading or fake online information can have serious consequences. It can affect the social, political, economic well-being of individuals, society, and also a nation as a whole. Fake News in the form of satire, fabricated and manipulated content, misleading information, conspiracy theories get more likes and shares on social media and they spread quickly in no time. Thus, fake news detection (FND) and prevention on social media platforms have gained tremendous attention among researchers. Fake news through online platforms poses unique challenges. Firstly, it is written intentionally and is subjective, making it very difficult to authenticate it based on news content. Secondly, social media information is unstructured and multi-modal, both aspects are complex to capture and integrate in fake news detection. Thirdly, fake information spreads very quickly and is mainly circulated through bots, trolls, and humans from varied backgrounds. Identifying such spreaders and victims is a challenging task. This article presents a critical review of literature on fake news detection. The state-of-the-art methods are discussed, most of the methods depend on news contents, user profiles, and social context features of a post. The importance of feature engineering, feature extraction and feature fusion in FND are highlighted. Various fake news detection datasets are discussed. Finally, future research directions in the fake news detection problem are presented.