wangziy4-3 / coinalysis

Blockchain graph analysis tools for Bitcoin. In collaboration w. Aurora Guo.

Coinalysis

Suite for analyzing blockchain data for address-based linking of shared send and centralized tumbler transactions.

Getting the Data

The first step to run Coinalysis is to download the blockchain data in its current form. One way to do this is to run blockchain_crawl/crawl_blocks.py after creating a results folder in blockchain_crawl. Then, move the resulting result subdirectories to final_analysis once processing is complete.

An RPC-enabled Bitcoin node is required for this.

A faster way is to download the data off an existing node (we provide one at http://162.248.4.26:5000) and unzip the ~40GB of blockchain data in the final_analysis folder.

The format of each result subdirectory is as follows:

• totaltime : total time spent processing these blocks by the script (may not be accurate if multiple script executions generated file)
• 1 folder per block. Each block folder has a simpletxs file, which has [transaction],[output amount] for each simple transaction in the block, and a file named for each transaction ID of all the complex transactions in the block

A simple transaction is defined as a transaction that has one input or two/one outputs (most transactions feature a transfer and change address).

Complex transactions are formatted like:

#input
[address] [amount]

#output
[address] [amount]

for example:

#input
1LkhppoM81Ni5Hb2rd54iiKuDS7rGUvAYj 0.02053905
169R5TLxsZ4PttgBcQceE7U32yoyT8gmYc 0.0418
1HmCnepma2Yvz7pomzc5gmDxBgSoo9EXN5 0.7357

#output
1HxtTp8yJ4e7PDgv9XtMZ6jqdFgy9a6BMh 0.3255
1Bwsj9BQHjr4Pdr11HqYL84bS5xQeFpMLm 0.0194
1Cw1BXVQsxCyrDz7JeLXpkJ3wjVZvkKP7T 0.45303905

for a three input three output transaction.

The invalidinputs and invalidoutputs in the top level folder have lists of transaction IDs where either the input or output was not parseable by the script (these transactions are also unable to be decoded by block explorers like blockchain.info).

The Web Interface

An RPC-enabled Bitcoin node is required for this.

The web interface is a script that allows you to input a transaction ID and view the possible flow decompositions inside that coinjoin transaction. An example instance is running at http://162.248.4.26:5000

To run the web interface, install Flask and run python run_webapp.py in the coinjoin directory.

Java is required for the backend graph partition worker.

Generating the Graphs

Generating the graphs for analyzing the anonymity sets of centralized tumblers proceeds by running the gather_data.py file followed by the draw_graphs.py file in the graphs folder. Move all Python files in this folder to the top level transaction graph storage folder before running.

Note of Warning

This is a class project :). Code is not intended to be optimized for maximum performance, readability, maintainability, (...). For example, the "parse_transaction" and "format_transaction" functions badly need refactoring, but are left as is because of technical debt we encountered about halfway through our data gathering. Pull requests to add comments and tests are also always appreciated.