System and method for generating virtual asset wallet address blacklist database based on graph attention network (GAT)

A system for generating a virtual asset wallet address blacklist database for virtual asset wallet addresses based on a graph attention network (GAT), wherein the blacklist database is utilized for a virtual asset transaction executed on a blockchain network through a transfer between a wallet address of a sender and a wallet address of a recipient among the virtual asset wallet addresses such that a virtual asset transfer amount is transferred from the wallet address of the sender to the wallet address of the recipient, and a transaction time and a virtual asset transaction detail including a transaction hash are recorded on the blockchain network, the system comprising:

• a GAT artificial intelligence (AI) engine server configured to train a GAT-based AI model using an index of a full node for each virtual asset pre-stored in an index database server, common transaction item information pre-stored in a virtual asset transaction analysis database server, and a main blacklist including high-risk virtual asset wallet addresses pre-stored in a main blacklist server, to calculate GAT scores based on the trained GAT-based AI model, to estimate the high-risk virtual asset wallet addresses using the calculated GAT scores, and to generate a GAT blacklist database including the estimated high-risk virtual asset wallet addresses;
• and
• a GAT blacklist database server configured to store the GAT blacklist database generated by the GAT AI engine server,
• wherein the index is generated by sequentially indexing the virtual asset wallet addresses,
• wherein the common transaction item information includes at least one of the wallet address of the sender, the wallet address of the recipient, the virtual asset transfer amount, the transaction time, the transaction hash, a previous transaction output, signatures for the virtual asset transaction, or any combination thereof,
• wherein the GAT AI engine server includes:
• a first data preprocessing module configured to preprocess the common transaction item information, label each of transactions based on the preprocessed common transaction item information according to each virtual asset, and query the transactions corresponding to a predetermined virtual asset wallet address;
• an AI learning module configured to perform GAT learning using the transactions queried by the first data preprocessing module;
• a teacher module configured to pseudo-label transactions that are unlabeled by the first data preprocessing module, and feed the pseudo-labeled transactions to the AI learning module to re-learn the pseudo-labeled transactions by data augmentation;
• and
• a first risk calculation module configured to calculate a risk level representing the GAT scores of virtual asset wallet addresses based on results of the GAT learning performed by the AI learning module,
• wherein the AI learning module extracts a GAT graph structure for the transactions based on the GAT using the index of the full node for each virtual asset and the common transaction item information,
• wherein the AI learning module generates a feature matrix by assigning a weight to each online source site during crawl collection for data matrixed by risk category using the main blacklist,
• wherein the AI learning module passes the feature matrix and an adjacency matrix through an attention layer to define nodes and edges of the GAT graph structure, such that a pre-trained model for labeled training data associated with the high-risk virtual asset wallet addresses is generated,
• wherein the AI learning module, using the GAT graph structure, predicts unlabeled data for the high-risk virtual asset wallet addresses and performs the GAT learning based on semi-supervised learning, such that the GAT scores of the virtual asset wallet addresses are generated,
• wherein the AI learning module automatically generates and updates the GAT blacklist database by repeatedly executing the GAT graph structure, the GAT blacklist database including the GAT scores, and
• wherein the GAT AI engine server is further configured to:
• receive data of an arbitrary virtual asset wallet address;
• obtain transaction data related to the arbitrary virtual asset wallet address;
• obtain a classified risk category for each transaction;
• generate an estimated probability value for each transaction based on prediction results generated by the GAT-based AI model;
• determine an overall risk level for the arbitrary virtual asset wallet address by combining the classified risk category and the estimated probability value for each transaction;
• and
• generate the GAT blacklist database based on the overall risk level.