Blockchain-Powered Deep Learning Model For Automated Bitcoin Entity Deanonymization And Forensic Analysis Of Transactions

Abstract

The pseudonymous design of the Bitcoin blockchain provides privacy to users but simultaneously enables illicit financial activities such as cybercrime, ransomware payments, and money laundering. This research proposes an advanced deep learning–driven framework for blockchain entity deanonymization and behavioural classification, aimed at strengthening transparency, security, and regulatory compliance. The primary purpose of blockchain in this research is twofold: first, as the immutable and distributed data source for large-scale transaction analytics, and second, as a trusted ledger for recording and auditing AI-driven classification outcomes. Building upon foundational machine learning approaches, the proposed system employs Elephant Swarm Water Search (ESWS) for optimizing tuned hyperparameters-tuned Attention-based Graph Neural Networks (Att-GNN) for a graph learning model for entity deanonymization, and to capture complex structural, temporal, and transactional dependencies among Bitcoin addresses and clusters. A labelled Bitcoin Entity Transaction containing 2,870 records has been collected from Kaggle. It trains supervised deep learning models to classify unknown entities, enabling automated representation learning directly from transaction graphs, eliminating reliance on handcrafted features. To enhance accountability, model predictions, confidence scores, and decision logs are stored on a private blockchain, ensuring tamper-proof audit trails for regulatory and compliance purposes. Experimental evaluations demonstrate significant improvements in classification accuracy 98.45%, precision 96.0%, recall 92.0%, F1-score 95.0%, and MCC 97.0%. This research contributes a secure and intelligent blockchain analytics solution that integrates deep learning with decentralized ledger technology to support automated risk assessment and forensic investigations. The ESWS-Att-GNN framework provides entity deanonymization and classification to enhance forensic analysis and auditability.