Performance Analysis of Deepfake Text Detection Techniques on Social-media

In the last few years, there has been remarkable progress in the domain of natural language generation & understanding. This has led to the development of enhanced text generation capability of machines that can generate spurious text or deepfake text. The technology may be exploited to change and shape public sentiment on social media. Therefore, mechanisms to detect deepfake text is a crucial task. The study emphasizes on detection of deepfake text on social media platform X (formerly Twitter). It presents the performance of machine learning classifiers Logistic regression (LR), XGB classifier, and AdaBoost classifier, utilizing feature engineering methods Term Frequency (TF) and Term Frequency-Inverse Document Frequency (TF-IDF). Deep learning methods such as Dense Neural Networks (DNN), Bi-directional Long Short-Term Memory (Bi-LSTM), and Convolutional Neural Networks (CNN)in addition to state-of-the-art models and variants such as Bi-Directional Encoder Representations from Transformers (BERT), DistilBERT & RoBERTa are also a part of study, leading to 89.17% accuracy in deepfake text detection.