Author: Thomas HUET 2022-11-08 14:22:41
Published on: 2022-11-08T14:22:41+00:00
In an email exchange between Thomas and Clara, they discussed the local reputation solution for Lightning Network (LN) and its potential effectiveness in mitigating attacks. They also talked about the need for a good formula to calculate reputation and how relying on "A satoshis per second" may not be wise. Clara mentioned other dimensions that need to be considered in the evaluation of any mitigation framework including centralization, protocol evolvability, and ecosystem impacts. She also suggested that the evaluation of effectiveness should consider protection and compensation for both attackers and victims. They talked about the need for an incentive-compatible scoring algorithm that incentivizes honest behavior from routing nodes. They also discussed the structure of the monetary and reputation strategies and how these could be improved to ensure fair compensation and accountability. Finally, they addressed some concerns around the unconditional fee policy and how it could be adjusted to account for time uncertainty and downstream payment path hops.The discussion on the Lightning-dev mailing list revolves around the issue of nodes failing to make payments. The upfront fees are rather low, so it is unlikely that avoiding "bad" nodes is primarily motivated by these fees. However, there is one recursive issue with the reputation mechanism: if a node assigns slots and liquidity to an upstream peer, such as Carol, this might constrain the incoming HTLC traffic of other honest peers, even if they each offer (K, L) of honest HTLC traffic. This is a bit of a "min-cut max-flow" problem for Carol, where the node's channel is her min-cut.There are also timing asymmetries in terms of high-reputation scores being earned during periods of low-congestion and consumed during periods of high-congestion. Therefore, reputation should be quantitative, rather than introducing a binary framework of low-risk/high-risk, to account for proportionality. However, this is a challenging issue, especially when addressing payments with intentionally delayed resolutions in a non-cooperative-only way. As reputation policy is specific to individual nodes, some nodes may choose to take congestion into account.Lastly, there is a conceptual issue with the chaining of unconditional fee and local reputation. The distinction between quick/slow jamming is based on the idea of maximal honest payment resolution delay. However, the bypass of this upper bound is only known after the HTLC forward, so the strategie regime (unconditional fee/local reputation) under which a HTLC forward should be evaluated is dependent on the knowledge of a future event. Despite this issue, reputation is used as part of a continuous game, evaluating a node's behavior over time rather than based on a single HTLC. Unconditional fees are always charged.
Updated on: 2023-06-03T10:36:25.205390+00:00