TL;DR
This paper introduces a Bitcoin-compatible protocol that allows intermediaries in a Payment Channel Network (PCN) to split payments locally and on-the-fly, drastically reducing transaction failure rates. By combining this with a novel routing algorithm called Interdimensional SpeedyMurmurs, the authors achieve a success ratio of over 90% for large payments, significantly outperforming the source-routing methods used in the current Lightning Network.
What Problem Does This Paper Address?
Payment Channel Networks (PCNs), like the Lightning Network, are the primary scaling solution for blockchains like Bitcoin. They allow users to transact "off-chain" instantly and cheaply. However, they suffer from a major reliability issue: high payment failure rates.
Currently, most PCNs rely on source routing, where the sender pre-selects the entire path to the receiver. The problem is that while the sender knows the network topology, they have no idea about the current liquid balances of channels they don't own. If a single link in the chosen path lacks sufficient funds, the entire payment fails. For large transactions, simulations show failure rates can be as high as 35-54%.
Methods
The authors propose a two-pronged solution:
- Local Splitting: Instead of forcing the sender to guess a perfect path, this protocol allows any intermediary node to split an incoming payment into multiple smaller "sub-payments" and route them through different neighbors. This happens dynamically based on the intermediary's real-time view of their own channel capacities.
- Interdimensional SpeedyMurmurs (INTSM): This is a new routing algorithm that builds on SpeedyMurmurs. It uses multiple spanning trees to calculate distances to the receiver. By looking across all these "dimensions" (trees) simultaneously, nodes gain much higher flexibility in choosing next hops while still guaranteeing loop-free routing.
- Unlinkability Extension: To prevent malicious nodes from "censoring" split payments (which they might do to save capacity for easier, non-split payments), the authors use additively homomorphic hashing and encryption (Paillier-based). This ensures that intermediaries cannot tell if two incoming sub-payments belong to the same larger transaction.
- Bitcoin Compatibility: Crucially, the protocol uses adaptor signatures, allowing these complex conditional payments to work on Bitcoin’s limited scripting language without requiring a hard fork.
Datasets
The authors evaluated their protocol using a mix of real-world and synthetic data:
- Lightning Network Snapshot: A real-world topology from March 2020 containing 6,329 nodes and over 84,000 channels.
- Synthetic Topologies: Scale-free (Barabási-Albert) and Random (Erdős-Rényi) graphs were used to test the algorithm's robustness across different network structures.
- Transaction Data: 100,000 simulated transactions with values following an exponential distribution (mimicking a mix of many small and few large payments).
Key Insights and Ideas
- Intermediaries Know Best: The "aha moment" is realizing that the sender is the least qualified person to pick a path. By moving the splitting decision to the intermediaries, the network can adapt to local liquidity fluctuations that the sender cannot see.
- The "Interdimensional" Edge: Standard routing often gets stuck if the "shortest path" is dry. By considering distances across multiple spanning trees at once, INTSM finds "good enough" paths that a single-tree approach would ignore, drastically increasing the pool of available liquidity.
- Unlinkability as Security: The paper argues that privacy isn't just for the user—it's a functional requirement. If nodes can identify split payments, they might drop them to avoid the risk of one part failing and locking up their collateral. Unlinkability forces nodes to treat all payments equally.
- Wormhole Protection: A side benefit of the homomorphic hash-lock approach is that it naturally protects against "Wormhole attacks," where malicious nodes try to steal routing fees from honest intermediaries.
Key Results
- Success Ratio: For payments with an average value of 25 EUR, INTSM with splitting achieved a >90% success rate, compared to only ~77% for Lightning-style shortest-path routing.
- Baseline Improvement: Even without splitting, the INTSM routing algorithm alone increased the success ratio by nearly 50% over standard protocols.
- Efficiency: The communication overhead remains low, generally requiring fewer than 15 "cPay" (conditional payment) operations per transaction.
- Strategic Splitting: The authors found that "strategic" splitting (prioritizing paths that are closer to the receiver) is far superior to random splitting, which actually performed worse than no splitting at all.
Limitations and Future Work
- Incentive Alignment (Fees): The current model abstracts away routing fees. Future work is needed to design fee structures that incentivize intermediaries to split payments rather than hoarding capacity.
- Metadata Privacy: While the protocol hides that a payment is "split," it does not currently hide the identity of the receiver from the intermediaries.
- Long-term Liquidity: The study focuses on the success of individual payments. The authors suggest investigating how local splitting affects channel depletion and "network health" over long periods of continuous operation.
Summary updated