# Cross-Chain Sealed-Bid Auctions Using Confidential Compute Blockchains

Crypto.world

> Sealed-bid auctions need bids to stay secret until the deadline, but public blockchains make everything visible, which exposes bidders to front-running and strategic inference.

**arXiv**: 2510.19491

**Published**: 2025-10-22

**Authors**: Jonas Gebele, Timm Mutzel, Burak Oez, Florian Matthes

## AI summary

## Overview

**Research area:** Blockchain privacy and decentralized finance — specifically sealed-bid auction design spanning a Trusted Execution Environment (TEE)-backed confidential compute chain and a public settlement chain.

**Technical level:** Intermediate. Readers need basic familiarity with smart contracts, escrow, and the idea of hardware enclaves, but the protocol itself is described in operational rather than mathematical terms.

**Scope:** The paper specifies, implements (on SUAVE with Ethereum Sepolia settlement), and evaluates a sealed-bid auction protocol that keeps bids confidential on a confidential compute blockchain while enforcing asset transfer and refunds on a public blockchain, without a trusted intermediary or multi-round bidder interaction.

## What This Paper Is About

Sealed-bid auctions need bids to stay secret until the deadline, but public blockchains make everything visible, which exposes bidders to front-running and strategic inference. The authors' goal is a protocol where bidders submit exactly one confidential, binding commitment, and where the winner is determined inside a hardware enclave and settled on a public chain that anyone can audit — with no trusted auction operator, no commit–reveal rounds, and no expensive cryptography.

## Key Contributions

1. **A cross-chain sealed-bid auction protocol** that offloads bid registration and winner determination to a confidential compute blockchain while escrow, ordering, and final settlement stay on a public blockchain, requiring only a single enclave call and a single settlement-layer transfer per bidder.
2. **A working prototype** built on SUAVE (Toliman testnet) as the confidential execution layer and Ethereum Sepolia as the settlement layer, published at `https://github.com/TimmMu/tee-confidential-auctions`, using the `confidentialInputs`, `confidentialStore`, `randomBytes`, and `doHTTPRequest` precompiles.
3. **A security evaluation** under stated TEE and confidential-compute-chain assumptions, mapping seven protocol goals (confidentiality, binding, non-malleability, enforceability, non-interactivity, auditability/verifiability, scalability) to the concrete mechanisms that realize them.
4. **A trust and scalability analysis**, including gas measurements per operation, a linear-cost resolution path, and a proposer-based resolution variant that reduces off-chain queries to one per proposal at the cost of temporary fund locking and lost non-interactivity.

## Main Findings

- **Bid confidentiality holds until resolution, but not after:** bids are encrypted to an attested enclave key and stored in sealed state; at resolution the enclave emits signed settlement transactions that reveal the participating funding addresses and transfers, enabling public auditability of the outcome.
- **Commitments are binding through escrow, not penalties:** each bidder receives an enclave-generated, per-bidder settlement-layer address and funds it directly; the protocol requires no collateral and applies no slashing.
- **Top-ups are permitted:** the protocol does not restrict bidders from increasing their bid before the cutoff; balances are evaluated at the pre-specified deadline block, so late transfers are disregarded.
- **Measured costs in the four-bidder experiment:** on Ethereum (L1) the operations cost 0 gas (deploy), 70,618 (start auction), 21,000 (submit bid, average), 0 (end auction), and 21,000 (claim valuable, average); on SUAVE the same operations cost 3,849,426 / 124,324 / 271,160 / 804,800 / 0 respectively.
- **Resolution cost grows linearly with bidder count** because each escrow balance must be verified through an individual off-chain query; per-bidder costs stay constant.
- **Gas figures for SUAVE are indicative, not final:** custom precompiles such as `doHTTPRequest` are priced at 1,000 gas units on the testnet, so the authors also report an adjusted scenario charging 100,000 gas per off-chain HTTP request.
- **The proposer-based resolution variant trades properties:** anyone can propose a candidate winner and the enclave checks only that the bid exceeds the current highest and that the address held sufficient funds at the cutoff block, cutting queries to one per proposal but temporarily locking funds and breaking non-interactivity.
- **Settlement-interface trust is a residual risk:** the enclave randomly samples n out of m endpoints from a publicly declared list and checks consistency, but a fully colluding subset could misreport balances or block heights; no cryptoeconomic deterrents are currently enforced.
- **Enclave compromise is a single point of failure:** the paper names Foreshadow, ÆPIC Leak, Plundervolt, and SGAxe as precedent attacks, and notes that a breach could expose settlement keys and the bidder-to-address mapping, enabling theft of escrowed assets and full de-anonymization.
- **Metadata leakage is only mitigated, not eliminated:** observers may correlate registrations with funding transfers via timing or address patterns; bidders are advised to randomize the delay between receiving an escrow address and funding it and to avoid reusing wallet identities across chains.
- **Comparison baseline:** the paper notes that the Riggs design costs approximately 2.5M gas per bid, whereas this protocol confines settlement-layer costs to standard asset transfers.
- **Liveness holds without active participation:** if no asset is escrowed the auction never opens; if no bids arrive the enclave returns the asset to the auctioneer; each signed settlement transaction has a beneficiary motivated to broadcast it and cover fees.

## Methodology in Plain English

The authors split the auction across two chains. A confidential compute blockchain runs the sensitive logic inside hardware-isolated enclaves; a public blockchain holds the money and the asset. The auctioneer sends the NFT (ERC-721) to an address the enclave generates, and the enclave publishes that address so anyone can confirm the asset is locked. A bidder then asks the enclave for a personal funding address, receives it encrypted under a key the bidder supplied, and simply transfers funds on the public chain — a transfer that looks like any other. The enclave keeps the bidder-to-address mapping in sealed state.

After the deadline, anyone can trigger closure, at which point the enclave queries public settlement-layer interfaces to confirm the deadline block was reached and to read each escrow balance at that block. It then signs Ethereum transactions paying the winner and refunding the losers, and emits the raw payloads for any relayer to submit. To test this, the team built a prototype on SUAVE's Toliman testnet (a go-ethereum fork with precompiles emulating enclaves, so TEE functionality is mocked via authenticated APIs) plus Ethereum Sepolia, compiled contracts with `solc` 0.8.20 and 10,000 optimizer runs, and measured gas in a four-bidder auction. They then analyzed what the protocol guarantees under stated adversarial assumptions and compared it against commit–reveal, cryptographic, and other TEE-based designs.

## Why This Matters

For research, the paper shows a way to get sealed-bid semantics without zero-knowledge proofs, MPC, or multi-round protocols, and it isolates the remaining trust in two places — the enclave and the settlement-layer data interface — rather than hiding it inside a centralized operator. It also gives measured gas numbers for both the confidential compute layer and the settlement layer, which prior TEE auction work (such as Trustee's centralized relay design) did not provide.

Real-world applications the authors point to:

- **Spectrum licensing auctions**, cited as a canonical first-price sealed-bid setting.
- **Carbon credit allocation**, explicitly named as a target DeFi use case for tokenized real-world assets.
- **Real-time digital advertising**, the motivating case where the paper cites Google's ad infrastructure generating over $260 billion in revenue in 2023 and antitrust findings from 2013–2019 about withheld second-price information, plus a 2018 arrangement with Facebook granting preferential bid access.
- **NFT auctions**, the concrete asset type the prototype assumes via ERC-721.

For industry, the appeal is deployability: the design needs no protocol modification, fits existing wallet tooling for encrypted responses (`eth_getEncryptionPublicKey`, `eth_decrypt`), and can use Flashbots bundles for gas sponsorship, so settlement fees can be decoupled from the enclave's transaction construction.

## Future Directions

- **Move settlement-layer interface logic into TEEs** to strengthen integrity, which the paper explicitly proposes as a way to reduce the residual trust in off-chain queries.
- **Add cryptoeconomic deterrents for interface misbehavior**, since misreporting is currently detectable and reputationally costly but not financially punished.
- **Harden against enclave compromise and side channels**, including key rotation, sealed-state re-encryption, and attestation freshness checks, which the paper discusses as open challenges for all TEE-based systems.
- **Re-deploy on a live confidential compute chain**: SUAVE has been discontinued, and the prototype's enclave guarantees are mocked via authenticated APIs on the Toliman testnet, so the protocol's claims remain to be demonstrated against real SGX-backed enclaves on an active platform.

## Target Audience

Researchers and protocol engineers working on blockchain privacy, DeFi market design, and auction mechanisms; practitioners evaluating TEE-based confidential compute chains (SUAVE, Secret Network, Oasis Sapphire, Phala Network, TEN Protocol) for production use; and anyone assessing whether sealed-bid price discovery can be brought on-chain without a trusted operator. Readers comparing privacy techniques — commit–reveal, zero-knowledge constructions, MPC, and hardware enclaves — will find the trade-off discussion and gas measurements most useful.

## Authors’ abstract

Sealed-bid auctions ensure fair competition and efficient allocation but are often deployed on centralized infrastructure, enabling opaque manipulation. Public blockchains eliminate central control, yet their inherent transparency conflicts with the confidentiality required for sealed bidding. Prior attempts struggle to reconcile privacy, verifiability, and scalability without relying on trusted intermediaries, multi-round protocols, or expensive cryptography. We present a…

[The full abstract on arXiv](https://arxiv.org/abs/2510.19491)

[Read the original paper](https://arxiv.org/abs/2510.19491)

---

Canonical: https://crypto.world/paper/cross-chain-sealed-bid-auctions-using-confidential-compute-blockchains-2510.19491
