Classifying Scam E-Commerce Shops with Supervised Learning

No Thumbnail Available

URL

Journal Title

Journal ISSN

Volume Title

School of Science | Master's thesis

Date

2024-09-20

Department

Major/Subject

Security and Cloud Computing

Mcode

SCI3113

Degree programme

Master's Programme in Security and Cloud Computing

Language

en

Pages

59

Series

Abstract

The global retail e-commerce industry has experienced significant growth in recent years. 20.1% of the retail purchases are expected to take place online in 2024. The industry rose from $5.8 trillion dollars market in 2023 to $6.3 trillion in 2024, which leads to an increase of 8.62% in just 1 year. Unfortunately, the scams targeting online shoppers are increasing at the same pace too. F-Secure is a global cyber security company, providing comprehensive protection to online life and securing digital moments. It specializes in consumer facing cyber security products and offers multiple products providing security benefits such as Browsing Protection, Trusted Shopping, Ransomware protection, Banking protection and many more. Primary goal of the thesis is on feature of Trusted Shopping. This thesis topic is aimed at extending machine learning capabilities of the verdict processing system of F-Secure i.e., Judge. The thesis first presents the core problem of the retail e-commerce industry particularly with the scam e-commerce shops where the customer ordered something but never received the product or sometimes with counterfeit goods, followed by answering how machine learning can be employed to efficiently classify the scam e-commerce websites. The thesis also discusses about the data collection and feature extraction processes that are inspired from previous researches. We conduct multiple experiments on the collected and extracted features with supervised learning algorithms such as Random Forest, Decision Trees, Extra Trees, XGBoost and compare the results. Finally, we learn from different metrics, several important features which possess similar patterns in most of the scam e-commerce shops. Random forest classifier indicated good results with 96.28% of accuracy in our study.

Description

Supervisor

Jung, Alex

Thesis advisor

Önen, Melek
Alnajjar, Khalid

Keywords

E-commerce websites, scam E-commerce websites, machine learning, classifiers, supervised learning, trusted shopping

Other note

Citation