Fraud Detection in Social Media: Representation Learning and Challenges

Research output: ThesisDoctoral Thesis

65 Downloads (Pure)

Abstract

Social reviews dominated the Web in recent years and have become a plausible source of product information. To promote specific products or defame others, fraudsters spread fake information through social review platforms. Effective detection of fraud reviews is not only important for consumer protection and fair trading, but also for review platforms' long term user retention. Different hot topics were highlighted in fraud review detection recently: bot review detection, due to the challenging nature of such review detection, the cold-start problem, crippling the fraud detection algorithm with new fraudsters introduced, and group fraudster detection, as such groups effectively manipulate people's tendency toward a product. Three approaches were proposed: a graph-based inductive learning to handle the cold-start problem, a framework based on Generative Adversarial Network to find fraud reviews, and a framework to model the spatial-temporal relations between reviewers in a group. Extensive experimental results demonstrated the effectiveness of the approaches.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • The University of Western Australia
Supervisors/Advisors
  • Togneri, Roberto, Supervisor
  • Liu, Wei, Supervisor
  • Bennamoun, Mohammed, Supervisor
Thesis sponsors
Award date10 Aug 2022
DOIs
Publication statusUnpublished - 2022

Fingerprint

Dive into the research topics of 'Fraud Detection in Social Media: Representation Learning and Challenges'. Together they form a unique fingerprint.

Cite this