Ανάπτυξη μηχανισμού ανίχνευσης κλοπής ταυτοτήτων στο Metaverse με χρήση μοντέλων τεχνητής νοημοσύνης

Μίχος, Γεώργιος

Development of an identity theft detection mechanism in the Metaverse using Artificial Intelligence models

Bachelor Dissertation

Author

Μίχος, Γεώργιος

Date

2025-09

Abstract

This thesis investigates Identity Theft detection in the Metaverse, combining a conceptual framework with a technical implementation. It first analyzes the key definitions (Metaverse, Digital Identity, Artificial Intelligence, Machine Learning) and situates the problem of identity theft within the broader domains of cybersecurity and privacy, highlighting why digital identity is a critical trust asset in the ecosystem of interconnected virtual worlds. The purpose and objectives are articulated both as a vehicle for informing and theoretical clarification (definitions, risks, the role of identities) and as an avenue for experimental investigation with ML, with a focus on Anomaly Detection. The study is organized into thematic sections (definitions/threat landscape, ML algorithms, evaluation criteria, system architecture, analysis of results) so that theory directly guides the design of the detection mechanism. On the technical side, an Identity Theft ML pipeline is implemented for data loading, preprocessing, training, evaluation, and visualization, and the algorithms Logistic Regression, Support Vector Machine, Random Forest, K-Nearest Neighbors, and Keras MLP are compared using accuracy/precision/recall/F1 metrics, with confirmation via cross-validation. Evaluation is conducted on two datasets: a behavioral one (User Activity) for avatar motion/pose profiling and a network-level one (UNSW-NB15) for attack detection. In the behavioral dataset, RF and KNN achieve excellent performance, demonstrating that simple, robust models suffice when behavioral patterns are clean and balanced. By contrast, on UNSW-NB15, binary detection is significantly more reliable than multi-class, where class imbalance and heterogeneity lead to performance degradation. The thesis concludes that Random Forest provides an excellent baseline for Identity Theft detection in the Metaverse, and it proposes future improvements through strengthening multi-class methodologies, targeted feature engineering, and trials on real platforms for near real-time monitoring.

Department

Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Πληροφορικής

Number of pages

Language

Greek

URI

https://dione.lib.unipi.gr/xmlui/handle/unipi/18253

Collections

Τμήμα Πληροφορικής

Show full item record

Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα