DETECTION AND ANALYSIS OF FRAUDULENT CONTENT ON WEB PLATFORMS

Abstract

Recently, the false information detection accompanying multimedia content entices numerous real-life applications such as election, emergencies, health care, terrorism, etc. One of the ultimate aims of artificial intelligence society is to develop an automatic system that can be recognized and understand fraudulent content accurately. Over the decade, many efforts made to recognize the false information accompanying multimedia data but still it is a challenging task to detect as sometimes sufficient evidence are not available to verify the content. To start with, we have reviewed the most popular and prominent state-of-the-art solutions, compared, and presented. Based on the literature survey, these solutions are categorized into handcrafted features-based descriptors and automatically learned features based on deep architectures. In this thesis work, the fraudulent content detection framework is divided into traditional machine learning(TML) and deep learning (DL) based architectures which are then utilized throughout this work. The first chapter detailed discussed about the technique employed for the prediction of fraudulent content having text as an input. An overview of the complete model is described in the following paragraph. The techniques we covered here in this concern are based on two ways. In the first case, the input is given as text embedded images, while in the other case, in the simple text format. In the text embedded images, using the OCR technique (optical character recognition) the text content is retrieved from an image. Whereas, the second case considered the text only content. These two cases have been considered in this chapter and techniques involved in each of these cases have been discussed in detail. In the next chapter, we considered the claim accompanying multimedia content (images and videos). Here, firstly we discussed the technique where the claim accompanies image content, and secondly, the technique concerning to the claim accompanies video content. The third chapter, elaborates the proposed multi-web platform framework for detecting deceptive claims on the social media platform. Spreading of misleading information on social web platforms has fuelled huge panic and confusion among the public regarding the Corona disease, the detection is of paramount importance. Previous studies mainly relied on a specific web platform to collect crucial evidence for the prediction of misleading information. The analysis identifies that v retrieving clues from two or more different web platforms gives more reliable prediction and confidence concerning a specific claim. This study proposed a novel multi-web platform voting framework that incorporates the 4 sets of novel features (including content features, linguistic features, similarity features, and sentiments features). To validate the claim, a unique source platform is designed to collect relevant headlines viz. YouTube and Google based on specific queries. The features are extracted concerning each collected headline. This unique platform can also help researchers to gather efficient headlines from various web platforms. After evaluation, it has been observed that our proposed intelligent strategy gives promising results and is quite effective in predicting misleading information. The model correctly detected about 98% of the COVID misinformation on the constraint Covid-19 fake news dataset. Furthermore, it is observed in our study that it is efficient to gather clues from multiple web platforms for more reliable predictions to validate the news. The proposed work provides practical implications for the policy makers and health practitioners that could be useful in protecting the world from misleading information proliferation during this pandemic. Finally, this thesis work is concluded with significant findings and future research aspects in the field of fraudulent content detection on social media.