Title: Probing Channel Balances in the Lightning Network

Abstract: As Lightning network payments are neither broadcasted nor publicly stored. Thus LN has been seen not only as scalability but also as privacy solution for Bitcoin. The protocol guarantees that only the latest channel state can be confirmed on channel closure. LN nodes gossip about channels available for routing and their total capacities. To issue a (multi-hop) payment, the sender creates a route based on its local knowledge of the graph. As local channel balances are not public, payments often fail due to insufficient balance at an intermediary hop. In that case, the payment is attempted along multiple routes until it succeeds. This constitutes a privacy-efficiency tradeoff: hidden balances improve privacy but hinder routing efficiency. In this work, we show that an attacker can easily discover channel balances using probing. This takes under a minute per channel and requires moderate capital commitment and no expenditures. We describe the algorithm and test our proof-of-concept implementation on Bitcoin's testnet. We argue that LN's balance between privacy and routing efficiency is suboptimal: channel balances are neither well protected nor utilized. We outline two ways for LN to evolve in respect to this issue. To emphasize privacy, we propose a modification of error handling that hides details of the erring channel from the sending node. This would break our probing technique but make routing failures more common, as the sender would not know which channel from the attempted route has failed. To improve efficiency, we propose a new API call that would let the sender query balances of channels that it is not a party of. We argue that combining these approaches can help LN take the best of both worlds: hide private data when feasible, and utilize public data for higher routing efficiency.