Overview

The transaction subgraph spiders aim at collecting the transactions related to the specific source addresses. BlockchainSpider gives the following strategies for sampling:

• BFS: Breath-First Search

• Poison: A kind of taint analysis technology.

• Haircut: A kind of taint analysis technology

• APPR: the Approximate Personalized PageRank algorithm.

• TTR: the Transaction Tracing Rank algorithm.

The performance comparison of the above strategies is as follows:

Chain Supports

The transaction subgraph spiders support to collect the transaction data of eth/bsc/polygon/heco until now.

• txs.eth: eth transaction subgraph spider.

• txs.bsc: bsc transaction subgraph spider.

• txs.polygon: polygon transaction subgraph spider.

• txs.heco: heco transaction subgraph spider.

The next sections will introduce more configurations about the transaction subgraph spiders through txs.eth.

Design with Graph Expansion

In BlockchainSpider, all transaction subgraph spiders are implemented based on the framework of graph expansion, which has four core operations:

• Expand: collects all transactions related to a given address.

• Push: merges the collected transactions to the subgraph.

• Rank: computes the relevance of addresses in the subgraph to the source address.

• Pop: select an address for expanding.

The implementation of different strategies are as follows:

Strategy	Push & Rank	Pop & Expand
BFS	Add expanded neighbors to a queue	Select a unexpanded node in the header of the queue for expanding
Poison	Add expanded out-degree neighbors to a queue	Select a unexpanded node in the header of the queue for expanding
Haircut	Allocated the pollution value of the expanded node to the neighbors according to the edge weight	Select a unexpanded node with the highest pollution for expanding
APPR	Use the local push procedure for extended neighbors	Select a node with the highest residual for expanding
TTR	Use the TTR local push procedure for extended neighbors	Select a node with the highest residual for expanding