AI Model Watches Bitcoin's Underworld for Illicit Transactions

A new study uses a hybrid artificial intelligence framework called E-GTNet to identify illicit transactions on the Bitcoin network. The model, developed by researchers at Southeast University in Nanjing and Peng Cheng Laboratory in Shenzhen, treats the Bitcoin ledger as a graph and incorporates temporal evolution features.

E-GTNet's edge-aware attention mechanism allows it to weigh the influence of neighbors differently, while its graph convolutional structure aggregates information across neighborhoods. The model also uses temporal sequence modeling to capture the dynamics of transactions over time.

The researchers validated the framework using the Elliptic Bitcoin transaction dataset and found that E-GTNet can identify anomalous nodes and structures in the network. The model's predictions were then visualized using a graph database, showing how value moves through illicit networks and how they evolve over time.