Paper 2025/1023

Universal Channel Rebalancing: Flexible Coin Shifting in Payment Channel Networks

Abstract

Payment Channel Networks (PCNs) enhance blockchain scalability by enabling off-chain transactions. However, repeated unidirectional multi-hop payments often cause channel imbalance or depletion, limiting scalability and usability. Existing rebalancing protocols, such as Horcrux [NDSS’25] and Shaduf [NDSS’22], rely on on-chain operations, which hinders efficiency and broad applicability. We propose Universal Channel Rebalancing (UCRb), a blockchain-agnostic, fully off-chain framework that ensures correct behavior among untrusted participants without on-chain interaction. UCRb incorporates the following core innovations: (1) a fair and reliable incentive-compatible mechanism that encourages voluntary user participation in off-chain channel rebalancing, (2) integration of Pedersen commitments to achieve atomic off-chain payments and rebalancing operations, while ensuring balance security, and (3) zero-knowledge proofs to enable privacy-preserving channel initialization and coin shifting, ensuring that user identities and fund allocations remain hidden throughout the process. We evaluate UCRb using real-world Lightning Network dataset and compare its performance against state-of-the-art solutions including Horcrux, Shaduf, and Revive [CCS'17]. UCRb exhibits a success ratio enhancement between 15% and 50%, while also reducing the required user deposits by 72%--92%. It maintains an almost negligible rate of channel depletion. Additionally, the long-term performance of UCRb is roughly 1.5 times that of its short-term performance, suggesting that continuous operation leads to improved efficiency. We implement a prototype for UCRb smart contracts and demonstrate its practicality through extensive evaluation. As \texttt{CoinShift} operations require no on-chain interaction, the protocol incurs minimal gas costs. For instance, opening and closing channels with 10 neighbors costs only 130K-160K gas—significantly lower than comparable solutions.