Motivation

Blockchain’s radical transparency, while valuable for auditability, has resulted in substantial privacy erosion. Studies by Makarov et al. (2023) estimate that 62 % of Ether transfers can be clustered to real-world identities through heuristic analysis alone. Mixers such as Tornado Cash offered partial relief but suffer from regulatory censure and composability dead-ends.  STELS seeks to restore user agency by enabling fine-grained, opt-in privacy primitives that integrate natively with the existing Ethereum execution environment.

Contributions

1. Formal Model. We define a composable privacy framework parameterised by unlinkability ε, front-run success probability p, and information leakage ℓ bits.

2. Protocol Design. We present STELS—a three-layer scheme marrying stealth addresses, relayer auctions, and zk-Gas Vouchers.

3. Reference Implementation. Full Solidity contracts and Circom circuits are released under MIT/Apache-2.

4. Evaluation. We benchmark gas costs, latency, and proof sizes on multiple L2 test-nets and perform security analysis under the Universal Composability (UC) framework.

Problem Statement & Objectives

Let A be a set of externally owned accounts (EOAs) on chain C. A transaction correlation function corr : A × A → {0,1} outputs 1 if two addresses are deemed linked by an adversary 𝒜 using publicly available data. The objective is to design a protocol Π such that for any PPT adversary 𝒜, the probability:

where ε ≤ 2⁻¹²⁸. Additional goals include:

• Front-run resilience: probability that an adversary can reorder an encrypted intent before execution p < 10⁻⁴.

• Fee unlinkability: mutual information I(sender; gasPayer) ≤ ℓ = ⌈log₂ gas⌉.

• Regulatory compliance: existence of a poly-time algorithm exportAudit(txid) that outputs a non-interactive proof binding txid to a user-controlled audit key.

Stealth Addresses & EIP-5564

First introduced by Buterin (2014) for early Ethereum privacy discussions and later formalised in EIP-5564 [4], stealth addresses leverage Elliptic Curve Diffie–Hellman (ECDH) to derive one-time recipient addresses. Unlike Zcash’s shielded pool or Monero’s sub-addresses, EIP-5564 is specifically crafted for Ethereum’s account model and gas constraints.

Account Abstraction (ERC-4337)

ERC-4337 [5] introduces UserOperation mempools enabling smart-contract wallets to define custom verification logic. STELS builds upon this by embedding zk-Voucher burn logic in the paymaster flow.

Network-Layer Privacy

Dandelion++ [6] mitigates timing correlation in Bitcoin by separating transactions into stem and fluff phases. STELS adopts a similar Poisson delay strategy at the relayer layer.

Zero-Knowledge Gas Obfuscation

Groth16-based SNARKs [7] offer succinct proofs with constant verification overhead. Tornado Cash and Zcash leverage similar constructs for value privacy; STELS repurposes them for fee privacy.

Research Methodology

This section documents the empirical and theoretical approach undertaken to derive the STELS design.

Threat Modelling Workshops

We conducted three formal STRIDE workshops with stakeholders (wallet vendors, compliance officers, and white‑hat MEV researchers) to enumerate adversarial capabilities. The resultant attack matrix informed parameter choices for ε, p, and ℓ.

Simulation Environment

Using go‑ethereum v1.29.0 patched with Flashbots SUAVE modules, we simulated 10 000 transactions across 128 virtual nodes. Packet traces were analysed via Wireshark to compute empirical deanonymisation rates.

Formal Verification

We modelled the protocol in the Tamarin Prover to establish unlinkability under the Decisional Diffie–Hellman (DDH) assumption and soundness of the zk‑Voucher construct. Proof sketches are included in Appendix D.

Benchmarking

Gas metrics were captured on Arbitrum Sepolia and zkSync Era test‑nets. Proof generation times were measured on an Intel i7‑13700H (12 cores) and Apple M2 Pro.

Cryptographic Foundations