Anolis: A Censorship-Resistant Anonymous Communication System with Zero-Knowledge Streaming Settlements and Asymmetric Rendezvous

Low-latency anonymous communication has long faced a set of structural contradictions: Tor-style networks rely on covertly distributed directories and unpaid volunteer nodes, making it hard to resist directory blocking and lacking sustainable node incentives; mix networks such as Nym and Lokinet, which maintain topology and incentives on public ledgers, mitigate global passive monitoring but expose gateways and relay endpoints to active censors' traversal blocking; and paying for transport service under the premise of anonymity means that the payment behavior itself may leave linkable traces between the communicating parties. This paper proposes Anolis, a peer-to-peer anonymous communication system driven by economic incentives and requiring no centralized services. The two communicating parties each build a multi-hop relay node chain and splice them into a bidirectional chained-proxy channel at a rendezvous point, with neither end knowing the other's network location or real identity; rendezvous signaling is anchored on blockchain consensus, exploiting the asynchrony of on-chain confirmation to decouple the two sides' connection-establishment timing and raise the cost of traffic-correlation attacks at the rendezvous phase; zero-knowledge streaming payment proofs hide payment contents and channel parameters while allowing either the payer or the relay to settle at any time, without trusted third-party arbitration; relay endpoints are not on-chain and can only be discovered through paid channels, raising the cost for attackers to enumerate nodes and launch traversal blocking; a pluggable transport layer mitigates deep-packet-inspection fingerprints. Paid service replaces the volunteer model: relays may provide service only after staking-based registration, and behavioral violations are absorbed by the guaranteed service fee and market reputation, so that relay supply sustains itself while nodes dynamically join and leave.