Cyber Informatics

Exploring signals of internal and external influence that contribute to the genesis and spread of online hate. The aim is to not only build models for predictive analysis of hate speech, but also counter-hate speech measures. This involves understanding the underlying social dynamics and network structures that facilitate the proliferation of such content.

Detection of fake news related to COVID-19. The aim is to overcome the existing challenges around this problem with limited data and early detection by leveraging unlabelled data and propagation patterns with deep learning models. This involves developing innovative algorithms to accurately identify misinformation and mitigate its spread.

Online users nowadays want to get quick visibility and following, hence try to explore shortcuts – one such shortcut is to approach the blackmarket services. These services are used to gain followers, likes, views etc through a third-party ecosystem. Our work focuses on proposing framework to detect these collusive users who are involved in suspicious activities on various social media platforms.

Misinformation spreads more rapidly on social media than true information. Therefore, it is essential to model how a certain type of misinformation spreads and what the driving factors are. This line of research aims at designing mathematical models to answer some of the pertinent questions related to misinformation

• How to predict the virality of misinformation

• How to design early-detection models to flag misinformation

• Who are the potential spreaders of misinformation and how we characterize such spreaders in terms of their personalities and values

• Can we identify the extent of the falseness of misinformation, etc.

Rapid increase of social media use entails rapid spread of misinformation in this age. It is exacerbated in pandemic like sitautions where people are vulnerable and information is scarce. In this project we perform a detailed analysis of claims and study the problems of unstructured online data. Given that we propose a claim detection model that would be adaptive to different domains, we suggest a definition of claim with greater scope. We extract and pre-process a large number of tweets containing claims to form the first large size annotated twitter dataset for this problem. Using our definition of claim, we also form a set of guidelines for future research in this domain. Detection of claims is followed by developing a model to evaluate its veracity using external resources such as knowledge graphs, web, fact checking websites, verified resources, external thesaurus, scientific papers and patents etc.