An Application of Random Walk on Fake Account Detection Problem: A Hybrid Approach

TL;DR

This paper presents a hybrid method combining graph-based and feature-based techniques, including SVM and SybilWalk, to effectively detect fake Facebook accounts with high accuracy on a large dataset.

Contribution

It introduces a novel hybrid approach integrating random walk, graph, and feature-based methods for fake account detection.

Findings

High accuracy achieved on Vietnamese Facebook dataset

Effective combination of graph and feature-based methods

Demonstrates the potential of hybrid approaches in social network security

Abstract

Social networks play a significant role in today's world. The importance of social networks, for example Facebook or Twitter, are undeniable. However, they also have many issues. One of which is the need for a defense mechanism against fake accounts. It is obviously not a trivial task to separate fake accounts from authentic ones. In this paper, we propose a ranking scheme, comprising of both graph based and feature based approaches to aid the detection of fake Facebook profiles. Utilizing Support Vector Machine (SVM) \cite{cortes1995} and SybilWalk \cite{JWZ17}, the model achieved high accuracy over the set of ten thousands Vietnamese Facebook accounts.