The Lightning Network Probing Simulator models channel balance probing in the Lightning network and measures the attacker's information gain and probing speed under various attack assumptions.
This software accompanies the paper "Analysis and Probing of Parallel Channels in the Lightning Network" by Alex Biryukov, Gleb Naumenko, and Sergei Tikhomirov. See also: blog post, slides, video presentation.
Requirements: python3, networkx, matplotlib.
Run run.py to print stats about the LN graph and launch two experiments based on the snapshot given in snapshots/. The first experiment measures information gain and probing speed for all parameter combinations (with / without jamming; direct / remote probing; simple / optimized probe amount selection). The results are saved as plots in results/. The second experiments measures information gain and probing speed for all configurations of two-channel hops (large / small channels, enabled / disabled in different directions). The results are presented as CLI output.
Usage example:
./run.py --num_target_hops=20 --num_runs_per_experiment=10 --min_num_channels=1 --max_num_channels=5 --use_snapshot --jamming
Run ./run.py -h for details.
The results in the paper were obtained as follows (running time approximately 1 hour):
./run.py --num_target_hops=20 --num_runs_per_experiment=50 --min_num_channels=1 --max_num_channels=5 --use_snapshot
./run.py --num_target_hops=20 --num_runs_per_experiment=50 --min_num_channels=1 --max_num_channels=5 --use_snapshot --jamming
For the 2021-09 version, the snapshot from 2021-09-09 was used. For the 2022-01 version, the snapshot from 2021-12-09 was used.