Identifying Machine Generated Tweets: Deepfake Detection on Social Media

In the digital age, the expansion of machine-generated content poses significant challenges to the integrity of online communications. This project addresses the critical issue of detecting fake tweets on Social networking websites platforms by utilizing modern advanced deep learning algorithms and fast text embedded data. Deep fakes, particularly machine-generated text, are becoming increasingly sophisticated, making it essential to develop robust methods for identifying such content. This approach integrates current deep learning models with fast text embedding techniques to differentiate between human-written and machine-generated tweets. These models are fine-tuned to classify tweets based on semantic and syntactic features captured through pre-trained embeddings. Specifically, quick text embeddings in conjunction with convolutional neural networks (CNN) and long short-term memory networks (LSTM) are used to propose a novel approach to deepfake detection using deep learning techniques in this study. Rapid text embeddings strengthen the model's capacity to distinguish between real and fake information by capturing semantic subtleties and contextual linkages in textual data. In order to extract hierarchical features from tweet sequences and to capture temporal dependencies, our approach combines these embeddings with CNNs and LSTMs. Using a dataset of tweets, the suggested method was assessed and found to perform better than conventional techniques in identifying content that was created by machines.