Website analysis: Finding Fraud Patterns in Web Pages

Tiago António Ratado Campos

Key information

Authors:

Tiago António Ratado Campos (Tiago Antonio Ratado Campos)

Supervisors:

Sandra Pereira Gama (Sandra Pereira Gama); Eurico José Teodoro Doirado (Eurico José Teodoro Doirado)

Published in

10/09/2020

Abstract

In the past decade, online advertisement fraud has been a growing issue, increasingly harming companies' business. There are several tools in the market that aim at protecting against advertisement fraud and, for this, Phybbit, a Japanese company, created SpiderAF. One of its unique features is the inclusion of a website blacklist that includes fraudulent websites. A website is classified as fraudulent after its web pages attributes and content were analyzed against know fraud patterns. In the ever evolving world of the web and advertisement fraud, new patterns keep emerging and its rapid detection is key to prevent fraud from impacting advertisers budget. We propose the application of Information Visualization techniques in order to create a system where researchers can look at the data from the several web pages and find fraud patterns, providing a more efficient alternative to what is available today. The newly discovered fraud patterns can then be used directly to analyze new websites, contributing to the protection of the digital advertisement ecosystem. We developed the system following an iterative and incremental design, closely with Phybbit. To validate our solution we conducted an usability evaluation and a utility evaluation done by Phybbit’s researchers. The system met the proposed objectives and showed promising results on both usability and utility evaluations.