FaNDS: Fake News Detection System Using Energy Flow

TL;DR

FaNDS is a novel fake news detection system that uses an Inconsistency Graph and Energy Flow to identify potentially fake news by analyzing the relationships and energy distribution among news items.

Contribution

The paper introduces FaNDS, combining graph-based inconsistency analysis with energy propagation to improve fake news detection accuracy.

Findings

FaNDS effectively identifies fake news with high sensitivity.

Compared to other methods, FaNDS shows improved detection performance.

High-energy nodes correlate strongly with fake news items.

Abstract

Recently, the term "fake news" has been broadly and extensively utilized for disinformation, misinformation, hoaxes, propaganda, satire, rumors, click-bait, and junk news. It has become a serious problem around the world. We present a new system, FaNDS, that detects fake news efficiently. The system is based on several concepts used in some previous works but in a different context. There are two main concepts: an Inconsistency Graph and Energy Flow. The Inconsistency Graph contains news items as nodes and inconsistent opinions between them for edges. Energy Flow assigns each node an initial energy and then some energy is propagated along the edges until the energy distribution on all nodes converges. To illustrate FaNDS we use the original data from the Fake News Challenge (FNC-1). First, the data has to be reconstructed in order to generate the Inconsistency Graph. The graph contains various subgraphs with well-defined shapes that represent different types of connections between the news items. Then the Energy Flow method is applied. The nodes with high energy are the candidates for being fake news. In our experiments, all these were indeed fake news as we checked each using several reliable web sites. We compared FaNDS to several other fake news detection methods and found it to be more sensitive in discovering fake news items.