The repository contains code to analyze the anonymity of different network-level anonymity enhancing schemes for cryptocurrencies by using a bayesian framework and different heuristics.

Different components

Dandelion

• The folder contains three files to analyze anonymity when varying parameters such as:
  • Forwarding probability (pf)
  • Network size (N)
  • Fraction of adversary nodes in the network (C)

Dandelion++

• This folder contains code for the same analysis as Dandelion i.e., varying pf, N and C. Along with that it contains code to:
  • Analyze anonymity when privacy subgraph is modified to be the bitcoin graph.

LN (Lightning Network)

• This folder contains code to perform analysis for:
  • An older 2018 LN topology snapshot
  • The most recent 2021 LN snapshot + analysis based on varying transaction amount
  • Longitudinal analysis of topologies from 2019 and 2020
  • Random graph topologies with balanced centrality are used instead of the existing centralized real LN topologies.
• All the folders consists of three code files:
  • Construct a directed graph with corresponding nodes and edge weights from a real LN snapshot (consisting of channel announcement and policy updates).
  • Perform anonymity analysis when we select nodes as asversaries with varying node degree.
  • Code to plot the data generated by the analysis scripts.
• There is also a seperate file for analysis when the best K paths are considered for routing payment instead of a single best path.
• The documentation about working of the code has been done for the 2018 topo files (other folders also have similar codes with minor differences).

Setup Details

• Dependencies can be installed using requirements.txt file.
• The codes are tested to work for Python 3.6 or higher.
• The codes for Dandelion and Dandelion++ are self sufficient. However, to run the codes for LN, data from real LN network snapshots is required, which is provided in the form of JSON files (see releases).
• LN code outputs data in different directories. This data can be used for plotting (using the plotting scripts provided in each folder) or to perform additional analysis.
• For LN, it would be great if the machine has multiple cores (20 or more). The code is already very parallelized to exploit the machine capabilities to the maximum. For normal laptop machines, running the simulator for larger topologies can be very time consuming. IN such a case it would be a good idea to start with the smaller 2018 topology.