roemba / thesis-rebalancing

Implementation of CoinWasher for my thesis at the Cybersecurity department of the TU Delft under the supervision of Dr. Z. Erkin and Dr. S. Roos

Network topology downloaded from https://explorer.acinq.co/

Running instructions

Running requires Gradle and Kotlin to work correctly. gradle run is sufficient to get everything executing. For now, to specify different scenario's, please comment/uncomment lines of code in src/main/kotlin/roemer/rebalancing/main.kt.

Dependencies

Running is currently only tested under Linux (Ubuntu 21.04) so your mileage may vary. All dependencies except for one are pulled in by Gradle.

If you wish to test out ReviveNode you need lp_solve to work correctly. To do this, follow only the 2. INSTALLATION instructions stated here and make sure to place your library in /usr/local/lib. Running ldconfig is not necessary as gradle run will already set the environment variable for you so it knows to search in /usr/local/lib. To complete the instructions, you need to download lp_solve_<version>_dev_<arch> and lp_solve_<version>_java via SourceForge for version ~5.5.

Code layout

The graphing model backing everything is based on JGraphT. An instance of this graph is held by ChannelNetwork to add PaymentChannel functionality to the graph. ChannelNetworks are instantiated via GraphHolder, which (for now) reads in a .txt or .json and runs the experiment.

A graph consists of Nodes, which hold a Queue that receives messages from other nodes. When a Node wishes to send a message to another Node, it deposits the message in the queue of the other node. The class ParticipantNodeAlt (an alternative, working implementation of the original ParticipantNode class) extends Node to add functionality for the participant discovery. In a similar manner, CoinWasherNode and ReviveNode extend ParticipantNodeAlt to add functionality of their respective algorithms on top of the participant discovery.

TODOs

• Debug CoinWasherNode
• Finish and debug ReviveNode
• Improve delays by using an actual latency model
• Generate balances and transactions using a model instead of using a random generator
• Setup proper testing
• Improve package structure, i.e. split roemer/rebalancing into model infrastructure and CoinWasher and split the Message.kt file into multiple, smaller files.
• Add execution independent time, i.e. event time

Logging

In case of large logfiles, pipe the output as follows:

gradle run > log.txt

and use less to display it using:

less -R log.txt

which will instruct less to also print the correct terminal colors