Novel approaches to fake news and fake account detection in OSNs: user social engagement and visual content centric model

With an increase in the number of active users on OSNs (Online Social Networks), the propagation of fake news became obvious. OSNs provide a platform for users to interact with others by expressing their opinions, resharing content into different networks, etc. In addition to these, interactions with posts are also collected, termed as social engagement patterns. By taking these social engagement patterns (by analyzing infectious disease spread analogy), SENAD(Social Engagement-based News Authenticity Detection) model is proposed, which detects the authenticity of news articles shared on Twitter based on the authenticity and bias of the users who are engaging with these articles. The proposed SENAD model incorporates the novel idea of authenticity score and factors in user social engagement centric measures such as Following-followers ratio, account age, bias, etc. The proposed model significantly improves fake news and fake account detection, as highlighted by classification accuracy of 93.7%. Images embedded with textual data catch more attention than textual messages and play a vital role in quickly propagating fake news. Images published have distinctive features which need special attention for detecting whether it is real or fake. Images get altered or misused to spread fake news. The framework Credibility Neural Network (CredNN) is proposed to assess the credibility of images on OSNs, by utilizing the spatial properties of CNNs to look for physical alterations in an image as well as analyze if the image reflects a negative sentiment since fake images often exhibit either one or both characteristics. The proposed hybrid idea of combining ELA and Sentiment analysis plays a prominent role in detecting fake images with an accuracy of around 76%.