BitPredict: End-to-End Context-Aware Detection of Anomalies in Bitcoin Transactions using Stack Model Network

The seamless integration of cryptocurrencies and blockchain technology in various sectors has revolutionized financial transactions. While cryptocurrencies serve as a convenient mode of payment, they have also opened avenues promoting fraudulent schemes such as Ponzi schemes, HYIPs, or money laundering activities leading to substantial financial losses. Traditional ways of anomaly detection, such as heuristic and signature-based approaches, have proven inadequate in addressing the intricacies of burgeoning fraud patterns. This paper explores the application of ensemble learning for anomaly detection in Bitcoin transactions by combining various ML techniques such as Isolation Forest, One-class SVM, and DBSCAN within a stacking framework. The proposed model harnesses the complementary strengths of each algorithm to achieve a nearly $98 \%$ accuracy rate in anomaly detection, thereby addressing the shortcomings of existing techniques. The study utilizes hyperparameter tuning techniques to enhance the effectiveness of the ensemble model and create a resilient model for detecting fraud and security threats in cryptocurrency transactions. Leveraging the cryptographic foundations of blockchain technology, the proposed method aims to create a more secure and reliable system for detecting threats and maintaining the integrity of Bitcoin transactions.