Fraud Detection in the Open Metaverse: Machine Learning Approaches for Secure Connected Communities

Document Type

Conference Proceeding

Source of Publication

Lecture Notes in Computer Science

Publication Date

8-2-2026

Abstract

This study aims to enhance the safety and trustworthiness of decentralized digital ecosystems by developing a machine learning framework for fraud detection in the Open Metaverse. Blockchain transactions—including sales, purchases, transfers, scams, and phishing—were classified using Random Forest, Gradient Boosting, K-Nearest Neighbors (KNN), Gaussian Naive Bayes, Decision Tree, and AdaBoost. Models were evaluated with accuracy, precision, recall, and F1-score, with Random Forest and Gradient Boosting achieving the highest accuracy of 97.9%, followed closely by KNN at 97.7%. To mitigate the imbalance between legitimate and fraudulent transactions, preprocessing steps such as SMOTE oversampling, information gain–based feature selection, and normalization were applied. The results demonstrate that both ensemble and lightweight models can effectively detect fraudulent activities while balancing predictive performance with computational efficiency. By integrating accuracy with fairness and efficiency, the study contributes to secure and responsible fraud detection practices for connected communities in the evolving digital economy.

ISBN

[9783032199805]

ISSN

0302-9743

Publisher

Springer Nature Switzerland

Volume

16462 LNCS

First Page

160

Last Page

172

Disciplines

Computer Sciences

Keywords

AI, Blockchain, Fraud Detection, Machine Learning, Metaverse

Scopus ID

105046286274

Indexed in Scopus

yes

Open Access

no

Share

COinS