The impact of a-star routing algorithm and channel un-blinding on bitcoin lightning network transaction success

The Lightning Network is a layer-2 payment channel network for Bitcoin to carry out transfer transactions faster and cheaper than on-chain transactions. LN network uses source routing to find path of transaction. But how LN finds the payment path is not always successful and efficient because the inbound and outbound balances of the payment channels are hidden. This generally causes transactions to fail or become more expensive. In this study, in order to increase transaction success, the effect of unblinding inbound and outbound balance of channels were researched. Addition to that, experiments were carried out, in order to understand the effect of using different path finding algorithms against common LN clients, which use modified Dijkstra algorithms. To study these effect analysis, Lightning Network nodes and transaction process simulators were used. Initially a simulator, using LND client's algorithm called CLoTH was utilized. xUsing this simulation system Martin's shortest path algorithm was implemented so that effects of multi criteria path finding were analyzed. Further throughout the research, a simulator which is easier to modify and being used in recent academic studies that is based on Python was being used. This simulator made it possible to analyze other common LN clients. (ie. Eclair, LDK etc.) By making necessary modifications on the simulator, A-star (A*) path finding algorithm was implemented. A heuristic for the A-Star algorithm was proposed and experiments were carried out with this. Moreover on a different branch, necessary modifications to un-blind channel balances were being done. Lastly, modification to leave channels as they are, but only a certain set of nodes to be un-blinded was implemented, and study was being completed. In the study, for the dataset, a snapshot of the LN as of 12th of April 2022, which was shared to literature by Rene Pickhardt was being used. For channel balances where necessary maximum one Bitcoin amount used randomly by binomial and uniform distributions. According to the study, there was no significant effect of using Martin's shortest path algorithm. Experiments carried out with A-star showed that, compared to Dijkstra based clients, for bigger amounts of transactions Dijkstra performed better. On the other hand, when run-time of simulation was considered, it was shown that A-star performed better. That meant for micro-transactions A-star could be used for more efficient transaction activities. In the experiments where channel balances un-blinded, high transaction success was being observed. On top of that, even though transaction amounts are increased, transaction success was not dropped significantly unlike a blinded network. For the experiments carried out when not all nodes were unblinded, it was shown that transaction success was significantly higher, which means that a sub-network, where privacy is not a big concern, can be built to increase overall network health.