Scrutinization of Text, Images and Audio Posts on Social Media for Identifying Fake Content

Social media platforms play a major role in dissemination of information across the world. The users of social media platforms like Twitter, Facebook, whatsapp etc. are exposed to enormous number of posts on various topics. It becomes infeasible for the users to decide the credibility of each message. As per Datareportal April 2023 global overview, there are 4.8 billion social media users equating to 59.9% of the world’s population. As per PEW Research Center, 2021 report “About half of the Americans get news from Social Media” as opposed to TV, news platforms, or the newspaper. The Fake content posted on Social Networks purporting to be authentic information has misleading effect on the readers. Therefore, it is very vital to detect false information posted on social media before it proliferates around the world. This paper addresses three types of fake content: Text, Image and Audio. This work presents a framework which is a fusion of three models to identify fake text, audio and images. Best techniques like Error Level Analysis, CNN for identifying fake images; Gensim, NLP for detecting phony Text content and Librosa, MFCC for detecting fraudulent audio content are identified and used. The proposed work achieved an accuracy of 94.9% for text, 87.0% for image and 99.8% for audio with minimum number of epochs.