Anomaly Detection in Open Metaverse Blockchain Transactions Using Isolation Forest and Autoencoder Neural Networks

The study explores anomaly detection in blockchain transactions within the Open Metaverse, utilizing Isolation Forest and Autoencoder Neural Networks. With the rise of the Metaverse, blockchain technology has become essential for secure digital transactions. However, the decentralized nature of blockchain makes it vulnerable to various anomalies, potentially undermining trust and security in digital spaces. Isolation Forest, an unsupervised machine learning algorithm, isolates anomalies based on the assumption that anomalies are few and distinct from regular data points. Its effectiveness in handling high-dimensional data makes it suitable for real-time applications. On the other hand, Autoencoders, a type of neural network, excel in detecting anomalies through reconstruction error, identifying data points that deviate from normal patterns. The research applied these models to a simulated dataset from the Open Metaverse, including features like transaction amount, login frequency, and session duration, to capture nuanced user behaviors. Preprocessing steps, such as one-hot encoding for categorical features and standardization for numerical features, ensured data consistency for accurate modeling. The Isolation Forest achieved a precision of 0.85, while the Autoencoder slightly outperformed it with a precision of 0.87. Both models demonstrated strong AUC-ROC values, with the Autoencoder scoring 0.85 compared to Isolation Forest’s 0.82, indicating robust performance in distinguishing normal from anomalous transactions. The findings underscore the potential of both models to enhance security in blockchain-based virtual environments, with the Autoencoder showing an edge in handling complex data patterns. However, the use of simulated data presents limitations, suggesting the need for further testing with real-world Metaverse transaction data. Future research could explore integrating other advanced algorithms, such as Graph Neural Networks, to improve anomaly detection in blockchain systems.