DETECTING DEEPFAKES ON SOCIAL MEDIA USING DEEP LEARNING AND FASTTEXT-BASED TWEET ANALYSIS

Abstract

Deepfake technology, which utilizes artificial intelligence to create manipulated media, poses a significant threat to the integrity of information on social media platforms. In India, the prevalence of deepfake content has surged, particularly within political and entertainment domains, where AIgenerated videos and fake news have gone viral, leading to widespread misinformation. This research aims to develop a robust AI-based model capable of accurately detecting deepfake content on social media, with a specific focus on identifying machine-generated tweets using FastText embeddings. Traditional detection methods relied on human moderation, fact-checking agencies, and manual filtering of posts using predefined rules and keyword matching. However, these approaches are timeconsuming, error-prone, and lack the scalability needed to manage the vast volume of online content in real time. Consequently, harmful and misleading information can spread rapidly before detection or removal. Given the growing impact of social media in shaping public opinion, this study is motivated by the need to combat misinformation and uphold the credibility of online discourse. Deep learning models offer a promising solution by automating the detection of deepfakes. In this approach, FastText embeddings are used to convert tweets into meaningful word vectors, which are then classified by deep learning models to determine whether they are human- or AI-generated. This method ensures real-time detection, enhanced accuracy, and greater scalability compared to conventional techniques.