Adaptive Multipath Proofs for Privacy Protection and Security in Payment Channel Networks

Payment channel networks enable scalable off-chain payments, but their practical deployment remains constrained by a persistent tension among routing efficiency, liquidity visibility, transaction privacy, and settlement security. Existing multipath routing mechanisms can improve payment success under fragmented liquidity, yet they often expose sensitive balance information, leak structural features of payment routes, and enlarge the attack surface for probing, channel exhaustion, and selective forwarding. This paper presents a novel framework, Adaptive Multipath Proofs (AMPs), for privacy protection and security in payment channel networks. The core idea is to bind multipath routing decisions with lightweight zero-knowledge verifiability, allowing intermediate nodes to validate path feasibility, fragment consistency, and settlement constraints without learning exact channel balances, the complete payment amount, or the global route structure. AMP integrates three mechanisms: a hidden-liquidity feasibility proof that supports privacy-preserving route selection, an adaptive payment-splitting strategy that dynamically determines fragment allocation according to network congestion and balance uncertainty, and a proof-coupled settlement guard that enforces atomicity and timeout consistency across all payment fragments. Together, these mechanisms reduce information leakage while preserving robust payment execution under dynamic network conditions. Experimental evaluation on real Lightning Network topologies and synthetic stress scenarios demonstrates that AMP significantly lowers balance disclosure and endpoint inference risk, improves payment completion under skewed liquidity distributions, and introduces only moderate computational and communication overhead. The results indicate that adaptive proof-carrying multipath routing offers a practical and effective direction for building secure, privacy-preserving, and high-success payment channel networks.