Fake news is a very dangerous problem for society, and its dangers have become clear in recent years, and research in this area is increasing, as evidenced by its impact on public opinion in the 2016 US presidential election. The dangers of fake news have social, political, and economic dimensions, and psychology also affects personality. This paper presents a solution to mitigate the impact of these messages. The system is designed to detect fake news and distinguish between them with less effort and less time. Most of smartphone users prefer to read social media news via internet. News websites publish news and provide authoritative sources. The problem is how to authenticate news and articles circulating in social media such as WhatsApp groups, Facebook pages, Twitter and other microblogging and social networking sites. Believing rumours and pretending to be news is harmful to society. Especially in a developing country like India, it takes an hour to stop the rumours and focus on the correct and authoritative news stories. This paper presents a model and methodology for detecting fake news. With the help of machine learning and natural language processing, we aggregate the messages and later try to use logistic regression to determine if the message is real or fake. The model works well and defines the accuracy of the results with 97.21% accuracy

Keywords : fake news, machine leaning, accuracy, authoritative sources.