A framework of fake news detection on web platform using ConvNet

Social media and web have become popular platforms of information sharing, knowledge gathering, expressing sentiments, perceiving choices regarding products and services through major news sources, and an active channel for marketing. Hence, with these promising features comes the threat of misinformation propagation, leading to undesirable effects. Therefore, accurate verification of fraudulent content on time is in high demand. Hence, to tackle this problem, we proposed a novel framework WSCH-CNN (web scrapping content heading CNN) model which counters the issue of fake (or mislead) news using convolutional neural networks (CNN). The framework consists of two CNN models named content model and heading model, which are used to find the linguistic similarities in fake news, and classifies them into real news or fake news. Both the models are evaluated on two publicly available datasets, namely Kaggle dataset and fake news challenge dataset, and two self-compiled real-world datasets, namely text dataset (text dataset of news articles) and multimedia dataset (Image dataset compiled from Facebook and Twitter), using accuracy, precision, recall, and F1 score as evaluation metrics. Moreover, the recognition accuracy achieved on these datasets is compared with similar state-of-the-art results. The proposed WSCH-CNN model proved quite successful in detecting the fake news with a high accuracy of 85.06% on multimedia dataset, 94.16% for heading model and 85.32% for content model which supersedes the state-of-the-art.