A Classification Algorithm to Recognize Fake News Websites

TL;DR

This paper introduces a logistic regression classifier that distinguishes fake news websites from reliable ones using website features, providing a reliability score to help users assess news credibility.

Contribution

The paper presents a novel classification algorithm and a dataset for identifying fake news websites, enabling automated reliability assessment.

Findings

The classifier achieved high accuracy in distinguishing fake from reliable news sites.

Website features like 'contact us' and security status are effective predictors.

The framework provides a quantifiable reliability score for news sources.

Abstract

'Fake news' is information that generally spreads on the web, which only mimics the form of reliable news media content. The phenomenon has assumed uncontrolled proportions in recent years rising the concern of authorities and citizens. In this paper we present a classifier able to distinguish a reliable source from a fake news website. We have prepared a dataset made of 200 fake news websites and 200 reliable websites from all over the world and used as predictors information potentially available on websites, such as the presence of a 'contact us' section or a secured connection. The algorithm is based on logistic regression, whereas further analyses were carried out using tetrachoric correlation coefficients for dichotomous variables and chi-square tests. This framework offers a concrete solution to attribute a 'reliability score' to news website, defined as the probability that a source is reliable or not, and on this probability a user can decide if the news is worth sharing or not.