Abstract
Payment Channel Networks (PCNs) like the Lightning Network (LN) currently compensate intermediaries only for successful payments. This creates a misalignment of incentives when payments fail, as intermediaries incur opportunity costs without compensation. This paper models these interactions using extensive form games and proposes a novel non-refundable fee mechanism. Simulations on real-world LN snapshots show that this mechanism increases payment success rates by 12.2%, reduces sender costs by 6%, and provides a robust economic defense against probing attacks.
Introduction: The Reliability and Security Gap
The Lightning Network relies on a series of intermediaries to route payments via Hash Time Locked Contracts (HTLCs). In the status quo, these nodes are only paid if the payment successfully settles. If a payment fails—due to liquidity issues, technical delays, or malicious activity—the intermediary receives zero compensation, despite having locked up capital that could have been used for other profitable routes.
This incentive structure leads to two critical vulnerabilities:
- Rational Refusal: Intermediaries may rationally refuse to forward payments if the expected utility (probability of success times fee) is lower than the collateral cost of locking funds. This can lead to a "reliability death spiral."
- Free Probing Attacks: Attackers can send intentionally failing payments to "probe" channel balances. Because failures are free, mapping the network's private liquidity carries no economic risk for the attacker.
Methodology: Modeling Incentives with Game Theory
The authors employ Extensive Form Games to characterize the sequential decision-making process of senders and intermediaries across a multi-hop path. In this model, nodes act based on their "beliefs" about the probability of a payment's success, aiming for a Bayesian Nash Equilibrium where every party maximizes their expected utility.
To address the identified gaps, the paper proposes a modified incentive model where a fraction of the routing fee is non-refundable. This ensures that intermediaries are compensated for the "collateral cost" ($c_i$) of locking funds, regardless of the final outcome.
The proposed model considers two primary cases:
- Guaranteed Compensation: The sender sets a non-refundable fee high enough to cover the intermediary's collateral cost regardless of the outcome.
- Expected Utility Alignment: The sender pays a lower non-refundable fee, relying on the intermediary’s belief that the total expected utility (including the potential success fee) remains positive.
Simulation and Real-World Data
The proposed mechanism was validated using a simulation built on a snapshot of the Lightning Network from February 2023. The dataset included approximately 15,000 active nodes and 60,000 channels. Realistic distributions for base fees, fee rates, and timelocks were derived from actual channel parameters to ensure the simulation reflected the network's true topology and economic environment.
Key Results and Insights
The shift to partial non-refundable fees yielded significant improvements across several metrics:
- Improved Reliability: The payment Success Ratio (SR) increased by 12.2%, and the Lock Ratio (LR)—a measure of intermediary willingness to participate—rose by 15.7%.
- The "Cheap Path" Paradox: While it seems counter-intuitive that adding a non-refundable fee would lower total costs, the simulation showed a 6% decrease in average routing fees for senders. By incentivizing reliable "cheap" paths to stay active, senders avoid falling back on expensive, high-fee alternative routes.
- Economic Deterrence: Probing attacks, previously free, now incur a linear cost. The study found that inferring 1 million satoshis of balance would cost an attacker approximately 1,500 satoshis.
Conclusion
By introducing non-refundable fees, the Lightning Network can align the incentives of senders and intermediaries more effectively. This transition transforms the network from a "success-only" payment model to one that recognizes the inherent cost of liquidity. While the paper proposes a modification to HTLCs to enable these fees, future work involves full protocol-level implementation and cryptographic proofs.
Summary updated