Use of Multiple Inputs and a Hybrid Deep Learning Model for Verifying the Authenticity of Social Media Posts †

Abstract

With the rise of social media platforms and the vast amount of text content generated on these platforms, text data forensics has emerged as a new area of research that aims to verify posts’ authenticity by analyzing textual content. This study proposes an innovative hybrid framework for detecting fake content on social media by examining both the text and metadata of Twitter posts. The metadata are fed into a feature selection method to select the most beneficial features. Using multiple inputs, a hybrid deep learning framework is proposed to classify Twitter posts as real or fake, where fake content is defined as posts containing misleading information. This research significantly contributes to the field of text data forensics by enhancing the detection of such fake texts. A recent comprehensive dataset for text data forensics called CIC Truth Seeker Dataset 2023 was used to assess the effectiveness of the proposed framework; the proposed framework uses long short-term memory (LSTM) to process textual data and hybrid residual neural network (ResNet) and deep neural network (DNN) layers for metadata. The framework has shown promising results during its preliminary evaluations. The paper examines the proposed model’s architecture and performance while highlighting potential improvements in privacy, ethics, real-time deployment, and implementation limitations to emphasize its broader impact.