Fake News Detection in Model Integral: A Hybrid CNN-BiLSTM Model

Abstract

The act of recognizing news that intentionally spreads false information via social media or traditional news sources is known as fake news detection. The characteristics of fake news make it difficult to identify. The spread of fake news and misleading information has increased dramatically due to social media's role as a communication tool and the quick advancement of technology. There is an urgent need for automated and intelligent systems that can differentiate between authentic and fraudulent information due to the fast dissemination of unverified content. The proposed hybrid model efficiently captures regional and worldwide relationships in textual details to address this by combining multiscale residual CNN and BiLSTM layers. The BiLSTM layers manage contextual representations and sequential dependencies, while the CNN layers concentrate on extracting deep local features. The model's capacity to recognize patterns of deception in textual content and comprehend semantic flow is enhanced by this dual architecture. The Edge-IIoT set data and the IoT-23 information from Aposemat were utilized in this study to assess the suggested framework empirically. A concept based on information transfer and sophisticated adaptive systems, we provide an understanding of outliers management paradigm of "generation–spread–identification–refutation" for identifying false information during emergencies. Findings from experiments clearly illustrate the superiority of the BiLSTM approach, demonstrating not only its state-of-the-art efficacy in identifying fake news but also its significant edge over traditional machine learning algorithms. This highlights the BiLSTM approach's critical role in protecting our information ecosystems from the ubiquitous threat of misinformation.