[Paper Review] Mutual Hyperlinking Among Misinformation Peddlers

Vibhor Sehgal, Ankit Peshin+2 more|arXiv (Cornell University)|Apr 20, 2021

Misinformation and Its Impacts36 references9 citations

TL;DR

This paper proposes a domain-level approach to identifying and demoting misinformation peddlers by analyzing mutual hyperlinking patterns among websites and Twitter users' sharing behavior. It finds that misinformation domains form tightly interconnected clusters with minimal links to trustworthy domains, and certain users consistently amplify these networks, enabling detection and potential demotion in recommendation systems.

ABSTRACT

The internet promised to democratize access to knowledge and make the world more open and understanding. The reality of today's internet, however, is far from this ideal. Misinformation, lies, and conspiracies dominate many social media platforms. This toxic online world has had real-world implications ranging from genocide to, election interference, and threats to global public health. A frustrated public and impatient government regulators are calling for a more vigorous response to mis- and disinformation campaigns designed to sow civil unrest and inspire violence against individuals, societies, and democracies. We describe a large-scale, domain-level analysis that reveals seemingly coordinated efforts between multiple domains to spread and amplify misinformation. We also describe how the hyperlinks shared by certain Twitter users can be used to surface problematic domains. These analyses can be used by search engines and social media recommendation algorithms to systematically discover and demote misinformation peddlers.

Motivation & Objective

• To address the growing threat of coordinated misinformation campaigns that exploit social media algorithms to spread lies and conspiracies.
• To develop scalable, domain-level detection techniques that complement per-post fact-checking, which is ineffective at scale.
• To identify structural patterns in hyperlink networks that distinguish misinformational domains from trustworthy informational ones.
• To uncover super-spreader Twitter users who systematically share links to misinformation domains.
• To provide tools for search engines and recommendation algorithms to demote entire misinformation networks based on structural and behavioral signals.

Proposed method

• Construct a domain-level hyperlink network by analyzing mutual hyperlinks between domains, using a Jaccard similarity threshold of 1% to define connections.
• Build a social media link-sharing network from Twitter user data, where edges represent shared links between domains by the same users.
• Apply community detection to identify clusters of domains with high internal connectivity, particularly those dominated by misinformational domains.
• Train a binary classifier to predict whether a domain is misinformational based on its hyperlinking patterns and Twitter user engagement metrics.
• Use a threshold-based approach to identify 'super-spreader' users who share links to multiple misinformation domains with high frequency.
• Leverage the identified structural and behavioral patterns to recommend demotion of entire misinformation networks in recommendation systems.

Experimental results

Research questions

• RQ1 Do misinformational domains exhibit significantly higher mutual hyperlinking compared to their links with informational domains?
• RQ2 Can the hyperlinking structure of a domain predict its likelihood of being a misinformation peddler?
• RQ3 Are there specific Twitter users who consistently share links to multiple misinformation domains, indicating super-spreader behavior?
• RQ4 Can the combination of hyperlink patterns and social media sharing behavior be used to train an effective classifier for identifying misinformation domains?
• RQ5 To what extent do misinformation networks form cohesive, isolated communities that are distinct from the broader web?

Key findings

• Misinformational domains exhibit significantly stronger mutual hyperlinking, with 89.3% of the largest community of 131 domains being misinformational, and 91.3% of a 46-domain community being misinformational.
• Misinformational domains have an average of 7.55 internal connections compared to only 1.57 external connections, while informational domains have 1.00 and 0.86 connections, respectively.
• A community of 14 climate-change denial domains, including wattsupwiththat.com, forms a highly interconnected network, with 13 out of 14 being misinformational.
• Certain Twitter users display predictable, high-frequency sharing patterns of links to multiple misinformation domains, indicating super-spreader behavior.
• The hyperlink structure alone can predict misinformation status with high accuracy, enabling scalable detection without per-post analysis.
• The study identifies a clear structural divide between misinformation and informational domains, with misinformation networks forming isolated, self-reinforcing clusters.