Ransomware Transaction Detection on the Blockchain with the TabNet Model

Ransomware is malicious software that encrypts a victim’s data, restricting their access and demanding a cryptocurrency ransom, typically in Bitcoin, for data release. The inherent privacy and intricate blockchain transaction process create challenges distinguishing between legitimate and ransom transactions. Consequently, a robust detection method is essential. Our approach involves a detection technique that analyzes various transaction features, including processed transactions, Bitcoin addresses, transaction weights, loops, lengths, counts, neighbors, years, and days. Leveraging the TabNet deep learning model and data from January 2009 to December 2018, our method outperforms traditional machine learning techniques, achieving a superior recall rate (0.61) and F1 Score (0.55) compared to decision trees’ 0.50 for both recall and F1 Score.