Determining an Optimal Threshold on the Online Reserves of a Bitcoin Exchange

This repository contains the source code for a paper on Bitcoin storage presented at the 15th Annual Workshop on the Economics of Information Security (WEIS 2016), held at UC Berkeley.

The key focus of the paper is to understand the behavior of a hot/cold Bitcoin wallet system which supports customer deposits and withdrawals, and suffers from periodic hot and cold wallet thefts, over a long time interval [0, T]. Specifically, we look at the net balance of the wallets at T as a function of the ceiling capacity of the hot wallet, with the goal of finding the optimal value for this threshold.

The code is divided into two directories - Theory and Simulations - each of which contains two fairly simple Java programs.

ExpBalanceTheoretical.java analytically computes the net balance as a function of various system parameters (e.g. deposit rate, hot wallet theft rate), provided as command line arguments. These results can be compared to those obtained via an event-driven simulation, ExpBalanceEventDriven.java, which computes the same value by repeatedly drawing the waiting times to and executing the next deposit, withdrawal, and hot wallet theft events in a loop, until a time counter reaches T.

The other pair of programs, ExpTimeTheoretical.java and ExpTimeEventDriven.java, output the expected time to an empty hot wallet. This value is used to determine the expected balance of the wallets, and is thus an intermediate result in our theory.

To compile and run any program, cd into the appropriate directory and run:

javac <filename>.java
java <filename> [arg1] [arg2] ...

For example, to run ExpBalanceEventDriven.java, execute the following from Simulations:

javac ExpBalanceEventDriven.java
java ExpBalanceEventDriven 80.0 78.0 0.01 0.01

To list the required and optional command line arguments that a program accepts, invoke the usage message, e.g.

java ExpBalanceEventDriven

Usage: java ExpBalanceEventDriven mean_d mean_w mean_t_h p_t_c [timespan iterations mu_low mu_high]
  e.g. java ExpBalanceEventDriven 80.0 78.0 0.01 0.01 20000 1000 100 130

For more information, please look at Sections 2 (Problem Formulation) and 5 (Approach) of our paper.

Authors: Samvit Jain, Edward Felten, Steven Goldfeder
Institution: Department of Computer Science, Princeton University