Abstract—Social media refers to communication channels on Internet that enable the creation and publication of content generated by the user and interaction between users. Given the accessibility to these means of communication and their rapidity, people resort more to them comparatively to the traditional media including radio, television and newspapers. However, dubious pieces of information such as fake news are often disseminated for malicious purposes. The proliferation of fake news has a strong negative impact on a society such as damage to the reputation of a personality, an organization or the aggravation of conflicts between its members. Due to the proliferation of fake news on these websites, the notion of veracity of information becomes a crucial issue. Research based on machine learning is promising. However, one of the main limitations is the efficiency of predictions. As a solution to detect fake news, we have proposed two models based on hybrid deep learning and evaluated our models on the two real datasets, namely ISOT and FA-KES. An experience of the proposed models to detect fake news, allowed to obtain on ISOT an accuracy of 99% for both models and on FA-KES, we obtain an accuracy of 68% for one the models and an accuracy of 63% for other. Other experiments in generalizing models on these data sets have proposed. The results obtained are better than other machine learning models.
 
Keywords—veracity, fake news, social media, artificial intelligence, machine learning, convolution neural network, recurrent neural network


Copyright © 2023 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.