Neural Fake Det Net - Detection and Classification of AI Generated Fake News

Abstract

Unreliable and deceiving information is spreading at a great speed these days across the world through social media sources. Fake news is a growing problem in our modern society, and it has become increasingly difficult to distinguish between real and fake news due to the advancement of technology. Fake or misinformation about the latest CORONA pandemic wreaked havoc. Studies conducted during the epidemic COVID that false news might have menaced public health broadly. Detecting and averting the spread of unreliable media content is a delicate problem, especially given the rate at which news can spread online. With the increase in the use of social media platforms; the leading cause for spreading such news can be that fake news can be published and propagated online faster and is also cheaper when compared to traditional news media such as newspapers and television. Online fake news or information which is deliberately designed to deceive readers is mostly commonly manually written; but with the recent progress in natural language generation techniques, models have been built to generate realistic looking ‘Fake news’. With the explosion of large language models fake news can be easily created and with proper grammar and sentences. This creates a greater need to handle the fake news identification problem in a different way to not just classify the fake and real news, but also to mark the human-generated and machine-generated (neural) fake news. Considering most of the work that is done in this research area, it is found that only the very complex language models that are used as generators and detectors are able to catch the machine generated fake news. Again, such models have been observed to be performing well on their own generated text, but not quite effective while working with text from other language models. Also, they don’t seem to be tested on the human generated fake news. Now if someone uses the language model to generate the news and then change a few elements manually to make it look more real; this kind of fake news might go completely undetected by such models. So, there is a considerable scope to further study and analyze the difference as well as similarities in the human and machine fake news. This study looks at the problem of machine-generated fake news classification as more of a comparative analysis of Human Vs Machine Generated fake news and identify the differences or similarities of the patterns.