This code introduces a set union algorithm to discern which groups of Bitcoin accounts are owned by the same entity. This concept was developed concurrently, in a much more formal and throughly executed form, by Dr. Sarah Meiklejohn as part of her PhD work at UC San Diego.
There are some interesting parts, and the set union idea was novel. However, this code was written by a 19-year-old research intern with no previous real-world coding experience. It's here more for historical purposes. I'm happy to help you work with it if you're interested. However, I've come a long way since then, and I make no promises about code quality.
This is an set of scripts used to analyze the Bitcoin blockchain. It operates on the output of the blockparser CSV call. It was written as part of a summer research project at Colgate University with my advisor Vijay Ramachandran.
The tools focus on parsing the CSV output into more usable format, grouping addresses by user based on multiple heuristics, and preparing the transaction data for network analysis.
To run all of the main scripts at once, first ensure that all four of the
blockparser CSV files (inputs.csv, outputs.csv, transactions.csv,
blocks.csv) are in this repo's directory. Then run bitcoindata. The new
data will be larger than the original files, and the original files will not be
deleted, so make sure you have enough space in the directory.
New files generated:
users.csv: A CSV file that associates addresses with their user IDs. This uses the classic user generation heuristic described by Reid & Harrigan.heurusers.csv: The same as above, but using the new heuristic, which also analyzes output data. A comprehensive description of this heuristic will be published soon.usersCount.csv: Associates each user ID with the number of addresses assigned to it, using the classic heuristic.heurUsersCount.csv: The same as above, but using the new heuristic.
We are using a machine with 22 GB of RAM, and some of these operations use all of it. Therefore, some of this code may be unrealistically slow on a typical personal computer.