PoS-CoPOR: Secure, Anonymous PoS Consensus

• PoS-CoPOR is a Proof-of-Stake protocol that embeds a native onion routing layer to anonymize leader identities and mitigate pre-proposal DoS attacks.
• It combines stake-weighted probabilistic leader election with ordered fallback leaders, ensuring genuine liveness even under adverse network conditions.
• Empirical evaluations show throughput of up to 110 tx/s with minor overhead from the anonymization process, balancing security and performance.

PoS-CoPOR is a single-chain Proof-of-Stake consensus protocol that mitigates the vulnerability of pre-elected leaders to Denial-of-Service attacks by integrating a native onion routing mechanism into the consensus protocol itself. Its design combines stake-weighted probabilistic leader election with an anonymization layer that conceals the network identity of the next block proposer, so that targeted DoS attacks on leaders before block production become more difficult. The protocol was implemented and evaluated with reported throughput of up to 110 tx/s with 6 nodes, while retaining the overhead of the anonymization layer within a modest performance range (Homoliak et al., 6 Oct 2025).

1. Protocol objective and system model

PoS-CoPOR addresses a specific tension in PoS systems: protocols that pre-elect leaders for subsequent rounds improve predictability and pipeline efficiency, but also expose those leaders to targeted IP-layer DoS before they can produce a block. The protocol therefore embeds the anonymization mechanism into the consensus path itself rather than treating privacy as an external network service. In the formulation provided for PoS-CoPOR, time is slotted, the chain is single-chain, and each round elects exactly one main proposer together with $A$ ordered alternatives (Homoliak et al., 6 Oct 2025).

The protocol’s network and adversarial model is explicit. It assumes an adaptive adversary controlling $\leq f$ fraction of stake, typically $f < 1/3$; the adversary may run Sybil nodes with tiny stake, eavesdrop on all P2P edges, and perform targeted IP-layer DoS. The network is partially synchronous, so honest messages are delivered within a known bound $\Delta$ except during rare partitions. This places PoS-CoPOR in a design space that is neither a classical BFT protocol with per-round voting finality nor a purely network-layer anonymity add-on. A plausible implication is that its main novelty lies in coupling proposer selection and transport obfuscation so that liveness is protected against adversaries that exploit leader foreknowledge.

2. Stake-weighted probabilistic leader election

Leader election in PoS-CoPOR is stake proportional and implemented over a fixed-size integer ring $[0,\mathrm{MAX})$, with $\mathrm{MAX}$ exemplified as $2^{32}$. Let $S=\sum_i s_i$ be the total active stake. Each node $i$ receives a contiguous interval of size

$$\mathrm{intervalSize}_i = \frac{s_i \times \mathrm{MAX}}{S},$$

with interval endpoints

$\leq f$0

Each round $\leq f$1 carries a 32-byte randomness value $\leq f$2, signed by the previous round’s leader. Selection proceeds by locating the current value within the ordered interval map; if the selected node has not already been chosen, it is appended to the leader list, and the process continues with $\leq f$3 until $\leq f$4 distinct leaders have been obtained. The first element is the main leader and the remainder are ordered alternatives (Homoliak et al., 6 Oct 2025).

The protocol additionally uses an $\leq f$5 parameter. After computing the ordered leader list, the first $\leq f$6 leaders are discarded and the returned pair becomes $\leq f$7. Setting $\leq f$8 yields the main proposer, while $\leq f$9 yields the first alternative, and so on. This ordered fallback structure is central to the liveness story: when a designated leader is offline or unreachable, timeout progression promotes alternatives without re-running an unrelated election.

Because interval assignment is stake weighted and deterministic given $f < 1/3$0, the election mechanism retains the usual PoS property that selection probability is proportional to stake. At the same time, the subsequent transport layer attempts to ensure that knowing which public key is likely to be selected does not reveal the proposer’s network location before block gossip begins.

3. Native onion routing layer

PoS-CoPOR embeds a lightweight, Tor-style onion layer into its P2P transport. Every node, upon joining, builds $f < 1/3$1 anonymization circuits of $f < 1/3$2 hops each, selected at random from the full node directory. The directory is static: all $f < 1/3$3 nodes and stakes are on-chain or in a directory service. For each circuit, the joining node picks $f < 1/3$4 distinct peers $f < 1/3$5, where $f < 1/3$6 is the exit, performs a key exchange for each hop, and shares symmetric keys $f < 1/3$7 (Homoliak et al., 6 Oct 2025).

Message encapsulation follows the standard onion pattern. For a raw message $f < 1/3$8, the sender computes

$f < 1/3$9

and then, for $\Delta$0,

$\Delta$1

The sender transmits $\Delta$2 to $\Delta$3. Each intermediate hop decrypts with $\Delta$4, learns only the next hop address and the inner payload, and forwards accordingly. The exit node decrypts with $\Delta$5, recovers $\Delta$6, and re-gossips it to transport peers.

Three anonymization modes are specified. In TOR-like, every gossip uses the full three-layer onion and the exit sends in clear to its peers. In Gossip-node, the first hop is onion and the exit then switches to mesh gossip under a long-term symmetric key, with no further onioning. In Dandelion, the first phase is the same as Gossip-node, but the exit broadcasts unencrypted to random peers, thereby breaking the path. These modes define different trade-offs between anonymity, path structure, and forwarding complexity.

The protocol does not claim that proposer identity remains permanently hidden. The block header carries only the proposer’s public key (coinbase), not its IP address. Once the block is gossiped, observers can read the header, learn the proposer’s public key, and map it on-chain to identity. The network address remains hidden until the proposer itself or its exit node broadcasts from a known socket. This suggests that PoS-CoPOR is designed for pre-proposal leader concealment and transport-level obfuscation rather than full post hoc anonymity.

4. Round execution, fallback leaders, and checkpointing

At the start of round $\Delta$7, each node initializes $\Delta$8 and invokes $\Delta$9. The procedure elects a leader using the previous randomness and the current alternative index. If the local node is the elected leader, it builds a block from the mempool, writes $[0,\mathrm{MAX})$0 into the header, signs the previous randomness, signs the block header hash, and gossips the block through the onion layer. Otherwise it sets a timeout $[0,\mathrm{MAX})$1 and waits. On timeout, $[0,\mathrm{MAX})$2 is incremented and leader selection is retried with the next alternative index (Homoliak et al., 6 Oct 2025).

Upon receiving a block $[0,\mathrm{MAX})$3, a node checks that $[0,\mathrm{MAX})$4, verifies the leader’s signature on $[0,\mathrm{MAX})$5, recomputes the elected leader from $[0,\mathrm{MAX})$6 and $[0,\mathrm{MAX})$7, and verifies that the proposer matches the designated leader for that alternative position. It then executes the block’s transactions, applies rewards to the leader and alternative leaders as specified, appends the block, and cancels the timeout.

No additional BFT voting is used per round. Finality instead emerges from chain quality and periodic checkpointing. Every $[0,\mathrm{MAX})$8 blocks, a $[0,\mathrm{MAX})$9-voting committee elected via VRF checkpoints the chain; after the checkpoint, no fork beyond that depth is accepted. The design is therefore explicitly described as trading off safety for liveness, similar to Nakamoto-style protocols. A common misconception is to classify PoS-CoPOR as a conventional BFT protocol because it includes a committee checkpoint. The protocol description does not support that reading: per-round agreement is chain-based, while the checkpoint is an auxiliary mechanism that bounds deep reorganizations rather than a round-by-round voting layer.

5. Security properties and attack model

The core defense is Leader Anonymity: the proposer’s public key is bound into the block header only at proposal time, so the adversary cannot learn the leader’s IP before the block is gossiped. The second defense is Circuit Hiding: each honest node gossips through an $\mathrm{MAX}$0-hop onion circuit, and deanonymization requires the adversary to control every hop in the path. The probability of a fully adversarial $\mathrm{MAX}$1-hop path is stated as

$\mathrm{MAX}$2

where $\mathrm{MAX}$3 is the adversarial node count. The third defense is Alternative Leaders: if the main leader is DoS’ed or offline, timeouts advance to $\mathrm{MAX}$4, preserving liveness under churn (Homoliak et al., 6 Oct 2025).

These mechanisms target a different threat than the classic stake-centralization literature. The stated attack surface is not compounding or unfair reward variance, but exposure of future proposers to targeted network disruption. The protocol therefore focuses on concealing network identity before proposal and maintaining a deterministic fallback chain of alternative leaders.

Several limitations are also explicit. There is a deanonymization risk if the adversary runs many low-stake nodes, because a node may accidentally construct a full adversarial circuit. The node directory is static and the peer set is fixed, so fully dynamic join/leave requires on-chain peer registration. The evaluation is limited to a local LAN environment, and future work is identified on geo-distributed deployment, analysis of $\mathrm{MAX}$5, and stronger formal modeling of anonymity under traffic analysis, including Dandelion++-style directions. Checkpointing relies on committee BFT, and future work is described as tightening finality time or exploring hybrid BFT/Nakamoto designs.

6. Empirical evaluation and performance envelope

The reported implementation ran on localhost with an Intel i5-9600KF, 32 GB RAM, and a Python 3 implementation. For throughput, the reported measurements with block size $\mathrm{MAX}$6 are as follows (Homoliak et al., 6 Oct 2025):

Nodes	Mode	Throughput (tx/s)
6	no-anon	$\mathrm{MAX}$7
6	TOR-like	$\mathrm{MAX}$8
6	Gossip-node	$\mathrm{MAX}$9
6	Dandelion	$2^{32}$0
6–2 offline	TOR-like	$2^{32}$1
9	no-anon	$2^{32}$2
9	TOR-like	$2^{32}$3
12	no-anon	$2^{32}$4
12	TOR-like	$2^{32}$5

The reported serialization and traffic costs are also concrete: a serialized transaction is approximately $2^{32}$6 B, a block header approximately $2^{32}$7 B, and a block with $2^{32}$8 transactions approximately $2^{32}$9 kB. Onion routing adds $S=\sum_i s_i$0–$S=\sum_i s_i$1 symmetric encryptions per message and raises per-block gossip by approximately $S=\sum_i s_i$2–$S=\sum_i s_i$3 traffic. The stated mode trade-off is that TOR-like offers the best anonymity with an approximately $S=\sum_i s_i$4 throughput hit, while Gossip-node and Dandelion provide slightly lower performance but simpler hops.

These measurements constrain interpretation. The data show that the anonymization layer is not free, but the observed penalty is materially smaller than would be expected from a design that delegated privacy to heavyweight external routing. At the same time, the reported environment is local rather than geo-distributed, so the quantitative balance between $S=\sum_i s_i$5, routing latency, and throughput remains an open deployment question.

7. Relation to earlier CoPOR usage and adjacent protocols

The label CoPOR is not unique in the PoS literature. In "Compounding of Wealth in Proof-of-Stake Cryptocurrencies" (Fanti et al., 2018), the term PoS–CoPOR is used for a framework centered on compounding, equitability, and reward-schedule design. That work models stake evolution as a time-inhomogeneous Pólya urn, defines equitability through the normalized variance

$S=\sum_i s_i$6

and proves that the geometric-reward schedule uniquely minimizes $S=\sum_i s_i$7 among honest PoS reward schedules with fixed total reward $S=\sum_i s_i$8. Its principal concern is wealth amplification under honest behavior and bounded strategic deviations, not network-layer DoS resistance.

In "Stability of shares in the Proof of Stake Protocol -- Concentration and Phase Transitions" (Tang, 2022), the provided exposition uses PoS-CoPOR to denote a “Proof-of-Stake Concentration and Phase-Transition” picture. There, investor shares $S=\sum_i s_i$9 converge almost surely, but their limiting behavior depends sharply on initial scale and reward schedule: large investors have stable shares, medium investors may be volatile, and small investors may shrink toward zero. Under super-geometric rewards $i$0 with $i$1, the exposition states that $i$2 almost surely and that exactly one investor eventually holds all the stake, a phenomenon termed “chaotic centralization.” This is a concentration-and-threshold analysis rather than a consensus protocol design.

A further adjacent line is "Resilient Consensus Sustained Collaboratively" (Chen et al., 2023), which introduces a Proof-of-Collaboration layer to protect existing PoS/BFT chains against long-range attacks. That protocol uses collaborative nonce search over disjoint SHA-256 slice assignments, miner attestations, slice shifts, and penalty certificates. Its objective is to harden committed blocks against long-range rewriting while avoiding wasteful PoW competition. It is therefore distinct from PoS-CoPOR’s pre-proposal leader concealment, although both works attach an auxiliary mechanism to improve a PoS system’s resistance to attacks that arise outside ordinary stake-weighted leader selection.

This terminological overlap matters. In current usage, PoS-CoPOR may refer either to a protocol with native onion routing for scalability and DoS-resistance (Homoliak et al., 6 Oct 2025) or, in earlier expository usage, to concentration/equitability frameworks concerned with compounding, dilution, and phase transitions (Fanti et al., 2018, Tang, 2022). The shared acronym does not imply shared mechanism.