Existing Decentralized Dark Pools

Workstream 2 of the decentralized crossing engine research. Reads against the vocabulary, design questions (Q1–Q5) and 7-axis rubric defined in 00-framing.md.

Scope

Surveys deployed and proposed systems that hide pre-trade trading interest and/or match it without a trusted operator: ZK-shielded DEXes, MPC dark pools, TEE confidential venues, encrypted-mempool / commit–reveal designs, and intent + solver batch auctions. Each is mapped onto the five privacy models from the framing doc and scored on the rubric. The deliverable distinguishes what runs in production from whitepaper claims — vaporware and not-yet-shipped features are flagged explicitly.

Out of scope (per framing): regulatory analysis, oracle vendor selection. Generic shielded-transfer protocols (Tornado-style mixers) are excluded — they hide transfers, not order matching, which is the crossing-engine problem.

Key questions

This workstream feeds the Phase-3 answers to Q1–Q5 with concrete prior art:

• Q1 (matching) — Which deployed systems do continuous crossing vs periodic batch auction, and what matching priority do they use over hidden orders?
• Q2 (privacy) — Which of the five privacy models has actually shipped a dark venue (pre-trade opacity, not just settlement privacy), and at what trust cost?
• Q3 (settlement) — Does any production privacy DEX settle cross-chain natively, and how (atomic, intent/solver, escrow/bridge)?
• Q5 (reference price) — How do crossing-style venues source a manipulation- resistant midpoint, and does any do it cross-chain?
• Cross-cutting — What is the gap between the "truly dark" marketing claim and the privacy that is cryptographically enforced and live today?

Findings

Renegade — MPC + collaborative-zkSNARK dark pool

Privacy model: MPC (with ZK). The only production system that is both pre-trade and post-trade dark by cryptography rather than trust.

• Privacy guarantee. Orders are secret-shared; relayers run a two-party MPC (VALID MATCH MPC) so neither relayer learns the counterparty's order unless it matches. Wallets are committed on-chain via hashes + nullifiers in a global Merkle tree; on-chain observers see only commitments. A "collaborative zkSNARK" wraps the MPC output in a ZK proof, so settlement is verifiable without revealing order data. [1][2][3]
• Matching mechanism (Q1). Continuous crossing network, functionally a CLOB-style book but encrypted. Orders are midpoint-pegged to the real-time Binance bid/ask — trades never cross a spread, so no slippage / price impact by construction. [4]
• Settlement (Q3). On-chain via Arbitrum Stylus contracts; the collaborative SNARK makes the match atomic — neither party can back out after the MPC. Live on Arbitrum One and Base. Each match settles on a single chain; no cross-chain leg support — different deployments are independent. [4][5]
• Trust / liveness. Permissionless P2P relayer gossip network; relayers are non-custodial and only get view access to compute pairwise MPCs. A relayer can decline to handshake (liveness depends on some honest relayer being online), but cannot steal funds or forge a match. Two-party-per-match MPC means a single uncooperative relayer just fails that match, not the venue.
• Throughput / latency. ~<1 s to match, <$1 gas per trade (Stylus). Per-pair MPC handshakes — throughput is relayer-bound, not chain-bound. [4]
• MEV exposure. Effectively zero pre-trade: no public mempool order, no visible size, fixed external midpoint removes the sandwich/front-run surface. Residual surface is relayer-side metadata (timing of handshakes).
• Maturity: live & proven. Mainnet on Arbitrum since 3 Sep 2024, also on Base. Note: a $209K exploit occurred and was resolved whitehat ($190K recovered) — a contract bug, not a privacy break. [5][6]

Penumbra — ZK-shielded chain with ZSwap batch-auction DEX

Privacy model: ZK-shielded chain + (planned) encrypted batch auction.

• Privacy guarantee. Penumbra is a Cosmos-SDK shielded chain: balances, transfers and LP positions are shielded notes (burn input note publicly, mint output note privately). Critical caveat: ZSwap's sealed-bid property — encrypting individual swap amounts — is NOT live. It requires "flow encryption" (homomorphic threshold encryption) which depends on Tendermint ABCI 2.0; until then individual swap amounts are publicly revealed before execution. Today only LP position open/close is anonymous; swap amounts are not dark. [7][8][9]
• Matching mechanism (Q1). Frequent batch auction: all swaps in a block clear together at a uniform per-block price against Uniswap-v3-style concentrated-liquidity positions. Batch interval = one block. [7]
• Settlement (Q3). Native on the Penumbra chain. IBC-connected to the Cosmos ecosystem — assets bridge in via IBC, but a single match does not span two chains atomically; cross-chain is asset import, not cross-chain crossing.
• Trust / liveness. Validators run the chain; once flow encryption ships they also control the threshold decryption keys (committee-trust).
• Throughput / latency. Per-block batching, no extra latency beyond Tendermint finality (~few seconds).
• MEV exposure. Batch + uniform clearing price removes intra-batch ordering MEV. But with amounts currently public pre-execution, front-running on block inclusion is still possible — the design's MEV defense is the not-yet-shipped flow encryption.
• Maturity: live but early. Mainnet since 2024; the dark-pool-defining feature (sealed-bid amount privacy) is paper/blocked-on-upstream, not shipped. Treat ZSwap-as-dark-pool as a roadmap claim.

CoW Protocol — solver-driven batch auctions (quasi-dark, MEV-protecting)

Privacy model: intent + solver batch auction. Not cryptographically dark.

• Privacy guarantee. Trust-based / weak. Orders are signed intents sent to an off-chain orderbook; they are visible to CoW's orderbook and to all solvers during the auction. There is no pre-trade opacity against the solver set or the operator — privacy is "not on a public mempool," not "hidden from the venue."
• Matching mechanism (Q1). Batch auction: intents are grouped into batches, solvers compete to produce the settlement maximising user surplus. Coincidence- of-wants (CoWs) are matched peer-to-peer inside a batch at uniform directed clearing prices (UDP); residual is routed to on-chain liquidity. [10]
• Settlement (Q3). On-chain settlement contract verifies the user signature and the order constraints (limit price, min-receive). Single-chain per settlement; no native cross-chain leg matching.
• Trust / liveness. Solvers bond COW, slashable for misbehaviour. The orderbook is operator-run (centralised liveness point). Solver set is permissioned-ish.
• Throughput / latency. Batch cadence ~seconds-to-minutes; very high real volume — $87B in 2025, >$9B in a single month — proves the architecture at scale. [11]
• MEV exposure. Strong protection against sandwiching and harmful reordering via batching + UDP; the solver still captures surplus and could in principle exploit visible flow (last-look-style), mitigated by competition + bonding.
• Maturity: live & proven as an MEV-protecting batch DEX — but it is not a dark pool: pre-trade interest is fully visible to solvers.

UniswapX — intent + filler Dutch auction

Privacy model: intent + solver. Not dark.

• Privacy guarantee. Weak / none pre-trade. User signs an off-chain Dutch order broadcast to a filler network; order parameters are visible to fillers.
• Matching mechanism (Q1). Per-order Dutch auction — price decays over time, first filler to accept wins. This is not crossing/batch; it is a competitive RFQ-style race per order. [12][13]
• Settlement (Q3). Signed off-chain order, on-chain settlement. Has a cross-chain mode: cross-chain UniswapX uses messaging-bridge settlement oracles to confirm execution; the production Uniswap cross-chain aggregator pairs UniswapX intents with Across for ~3 s L2↔L2 / ~30 s mainnet→L2 fills across 9 networks. This is the most mature cross-chain intent settlement in the survey. [12][14]
• Trust / liveness. Filler network competes; relies on at least one filler bidding. Cross-chain leg adds bridge/oracle trust.
• MEV exposure. Dutch decay leaks the order to fillers; fillers can extract the gap between decay price and true market. No pre-trade opacity.
• Maturity: live & proven for intent settlement incl. cross-chain; not a dark venue.

1inch Fusion / Fusion+ — intent + resolver Dutch auction

Privacy model: intent + solver. Not dark. Architecturally near-identical to UniswapX.

• Privacy / matching. Signed order intent with a price range + time window; professional resolvers compete in a Dutch auction (rate decays from desired to minimum-return). Orders visible to resolvers. [15]
• Settlement (Q3). On-chain; resolvers bundle txs to avoid the public mempool. Fusion+ extends to cross-chain swaps without bridges, using a hashlock/escrow (HTLC-style) atomic-swap construction across chains — relevant prior art for Q3 atomic cross-chain settlement. [15]
• MEV exposure. Resolver bundling gives front-running protection at inclusion time; resolver still sees the order. Not pre-trade dark.
• Maturity: live & proven, billions/month combined with UniswapX/CoW; Fusion+ cross-chain live. Not a dark venue.

Shutter Network — threshold-encrypted mempool

Privacy model: encrypted mempool / commit–reveal (threshold).

• Privacy guarantee. Cryptographic but temporary and shallow. A Keyper committee runs DKG; users encrypt txs to the committee public key; the proposer orders ciphertexts; after inclusion the committee publishes decryption shares and the tx executes. Hides tx content until ordering is fixed — it does not provide a persistent dark order book, and metadata (gas price, sender address, tx size) leaks and can itself enable attacks. [16][17][18]
• Matching mechanism. None of its own — Shutter is mempool infrastructure, not a matching engine. A dark venue would have to be built on top.
• Settlement / cross-chain. Inherits the host chain; no cross-chain matching.
• Trust / liveness. Threshold committee — honest-majority of Keypers for liveness and privacy; committee collusion breaks privacy. Small live Keyper set today → longer inclusion times.
• Maturity: live but early. First mainnet threshold-encrypted mempool on Gnosis Chain; useful as a front-running shield primitive, not a dark pool.

TEE-based designs — SUAVE, Oasis Sapphire, Tristero, academic cross-chain

Privacy model: TEE (hardware-trust).

• Flashbots SUAVE. TEE coprocessor concept: bundles encrypted to a known SGX measurement, enclave decrypts/orders, block sealed before any operator sees contents. suave-geth was archived (read-only) on 12 May 2025 — the project is effectively discontinued. Flag as dead-end / vaporware for build purposes; the idea (enclave-ordered confidential bundles) survives in the TEE literature but not as shippable Flashbots product. [19]
• Oasis Sapphire. Confidential EVM ParaTime: standard Solidity runs inside Intel SGX; calldata, state and storage encrypted; node operators cannot see inputs/outputs. ~6 s finality. Live as a confidential-EVM chain — a private order book could be deployed on it as an ordinary contract, but Sapphire itself ships no dark-pool/matching application. It is a confidential substrate, not a venue. [20]
• Tristero. Explicitly a "truly dark" dark pool: encrypted order intents delivered to a network of Intel SGX TEE nodes running an orderbook, account- based model, ZK + TEE. Funded ($4.8M seed, 2023) and frequently cited in taxonomies, but no evidence of a live mainnet product — treat as testnet / paper until a deployment is verifiable. [21][22]
• Academic cross-chain sealed-bid auction (arXiv 2510.19491, 2025). Research prototype: sealed-bid auctions across chains using TEE/confidential-compute blockchains; bid amounts, bidder identity, interim outcome hidden; trust = TEE attestation + confidential-compute infra. Research prototype only (PoC code, not production) — but it is the closest published design to the cross-chain crossing problem and confirms the TEE route is technically the shortest path to cross-chain privacy. [23]
• General TEE caveat. Privacy is hardware-vendor trust (Intel SGX/TDX); side-channel and enclave-break history means TEE-only privacy is not cryptographic-grade. Low latency is the upside; "trust the chip" is the cost.

Penumbra-adjacent / Aztec / Starknet — ZK-shielded substrates

Privacy model: ZK-shielded.

• Aztec Network. Privacy-first ZK L2 on Ethereum with private smart contracts and selective visibility inside the app layer. The a16z "cross-chain sealed-bid auction using Aztec Connect" demo showed account anonymity for auctions (same account reused across auctions without linkage). Aztec Connect was sunset; the current Aztec L2 (Ignition chain live Nov 2025, token Dec 2025) is a general programmable-privacy L2 — a dark pool would be an app to build, none ships today. Substrate, live but early; no dark venue product. [24]
• Starknet confidential DeFi. strkBTC brings shielded balances / confidential transfers to Starknet — confidential transfers, announced/ roadmap, not a matching venue. Paper/early.
• These are ZK substrates: they could host a dark pool the way Sapphire could, but none provides matching.

Railgun & Panther — shielded-DeFi privacy systems

Privacy model: ZK-shielded.

• Railgun. On-chain ZK privacy system on Ethereum/BSC/Polygon/Arbitrum;

  $100M shielded, ~$3.5B private cumulative volume; EF staked 50K RAIL. Live & proven — but it shields interactions with existing DeFi (private balances, private calls into AMMs); it is not a matching engine or dark pool. Relevant only as a privacy-wrapper precedent. [25]

• Panther Protocol. Privacy-DeFi with a "compliant" angle; reached Limited Mainnet Beta (Canary) on Polygon Mar 2025, Veridise audit, codebase open-sourced May 2025. Taxonomies list it with an orderbook ambition, but the live product is a shielded-pool/compliance system, not a proven dark-pool matching venue. Live but early; dark-pool matching unproven. [26]
• Singularity (DarkSwap). Institutional on-chain dark pool: ZKPs prove order legitimacy/compliance without revealing asset/price/size; FHE lets a "book node" match encrypted orders without decrypting. Architecturally the most ambitious ZK+FHE dark-pool design surveyed. Marketing presents it as operational, but FHE-based encrypted matching at production latency is unproven and no independent volume/audit evidence was found — treat as testnet / early, claims unverified. [27][28]

Comparison table

Rubric axes: Privacy guarantee · Trust assumptions · Liveness/censorship · Throughput/latency · Cross-chain · Maturity · MEV exposure. PM = privacy-model classification.

System	PM	Privacy guarantee	Trust assumptions	Liveness / censorship	Throughput / latency	Cross-chain	MEV exposure	Maturity
Renegade	MPC+ZK	✅ Crypto pre+post-trade dark; size hidden	⚠️ Honest relayer must be online; non-custodial	✅ Any honest relayer suffices; permissionless	✅ <1 s match, <$1 gas	❌ Single-chain per deploy (Arb, Base separate)	✅ ~Zero pre-trade (fixed midpoint, no public order)	✅ Live & proven (Arb Sep 2024)
Penumbra ZSwap	ZK-shielded + (planned) enc. batch	⚠️ LP anon live; swap-amount privacy NOT live	⚠️ Validator set; future threshold committee	✅ Chain-level	✅ Per-block batch	⚠️ IBC asset import, not cross-chain match	⚠️ Batch removes ordering MEV; amounts public today	⚠️ Live but early; dark feature paper
CoW Protocol	Intent + solver batch	❌ Orders visible to solvers/operator	⚠️ Operator orderbook; bonded solvers	⚠️ Operator-run orderbook	✅ $87B/yr at scale	❌ Single-chain settlement	✅ Strong vs sandwich (UDP); solver surplus residual	✅ Live & proven (not a dark pool)
UniswapX	Intent + solver	❌ Order visible to fillers	⚠️ Filler network; bridge oracle for x-chain	⚠️ Needs ≥1 filler bidding	✅ ~3 s L2↔L2	✅ Native x-chain (Across, 9 nets)	⚠️ Fillers see/extract Dutch-decay gap	✅ Live & proven (not dark)
1inch Fusion+	Intent + solver	❌ Order visible to resolvers	⚠️ Resolver set; HTLC escrow x-chain	⚠️ Needs ≥1 resolver	✅ Seconds	✅ Bridgeless x-chain (hashlock/escrow)	⚠️ Resolver bundling shields inclusion only	✅ Live & proven (not dark)
Shutter Network	Encrypted mempool	⚠️ Content hidden till ordering; metadata leaks	❌ Honest-majority Keyper committee	⚠️ Committee liveness; small set today	⚠️ Added inclusion latency	❌ Host-chain only	⚠️ Blocks front-run pre-inclusion; not persistent dark	⚠️ Live but early (Gnosis); no matching engine
Flashbots SUAVE	TEE	⚠️ Enclave-sealed bundles (concept)	❌ SGX vendor + enclave integrity	n/a	n/a	n/a	(concept)	❌ Discontinued — repo archived May 2025
Oasis Sapphire	TEE	⚠️ Inputs/state/output hidden from operators	❌ Intel SGX/TDX hardware trust	✅ Chain-level	✅ ~6 s finality	❌ Single chain	⚠️ Confidential substrate; depends on app	⚠️ Live substrate; no dark-pool app ships
Tristero	TEE (+ZK)	⚠️ Claimed "truly dark" via SGX node net	❌ SGX vendor + TEE node honesty	unknown	unknown	unknown	(claimed dark)	❌ Testnet/paper — no verifiable live product
Academic x-chain sealed-bid (2510.19491)	TEE	⚠️ Bid/identity/outcome hidden	❌ TEE attestation + conf-compute infra	n/a	n/a	✅ Designed cross-chain	(by design dark)	❌ Research prototype (PoC only)
Aztec Network	ZK-shielded	✅ App-layer programmable privacy	⚠️ ZK L2 validator set	✅ Chain-level	⚠️ L2 latency	❌ Single L2	⚠️ Substrate; depends on app	⚠️ Live but early substrate; no dark venue
Railgun	ZK-shielded	✅ Shielded balances/calls (crypto)	✅ Trustless ZK on host chains	✅ Chain-level	✅ Production scale	⚠️ Per-chain deployments	⚠️ Privacy wrapper, not a venue	✅ Live & proven (not a dark pool)
Panther Protocol	ZK-shielded	⚠️ Shielded pool; "compliant" privacy	⚠️ Audited; compliance gating	⚠️ Early mainnet beta	unknown	❌	⚠️ Not a matching venue	⚠️ Live but early; dark matching unproven
Singularity DarkSwap	ZK + FHE	⚠️ Claimed crypto-dark (asset/price/size)	⚠️ Distributed keygen; FHE book node	unknown	❌ FHE-match latency unproven	unknown	(claimed dark)	❌ Testnet/early; claims unverified

Implications for our engine

Viable foundations

1. Renegade's MPC + collaborative-zkSNARK is the only deployed design that is genuinely dark (Q2) and genuinely decentralized at once. It proves cryptographic pre-trade opacity and atomic settlement can run in production at <1 s. Two architectural patterns are directly reusable for our engine:

   • Midpoint-peg crossing (Q1 + Q5). Pegging matches to an external reference (Binance midpoint) sidesteps the hardest dark-pool problem — price discovery over hidden orders — and eliminates spread/slippage MEV by construction. Our cross-chain engine should similarly import a reference price rather than discover one internally.
   • Collaborative SNARK = atomic match (Q3). MPC that outputs a ZK proof means the match is non-repudiable before settlement. This is the right primitive for a two-leg cross-chain settlement: prove the match, then settle each leg.
   • Gap to close: Renegade is single-chain. Our engine must add the cross-chain settlement leg Renegade lacks.

2. Intent + solver settlement (UniswapX, 1inch Fusion+) is the proven cross-chain settlement layer (Q3) — but not the privacy layer (Q2). Fusion+'s bridgeless hashlock/escrow atomic swap and UniswapX's bridge-oracle model are the only live, at-scale cross-chain settlement mechanisms in the survey. The viable architecture is a hybrid: do the dark matching with a Renegade-style MPC/ZK core, then hand the matched cross to an intent/solver- style settlement layer for the cross-chain legs. The framing doc's Q3 "intent- settled with solver/filler capital" is the realistic path; pure atomic cross-chain swap exists (Fusion+) but constrains asset types to those an HTLC can lock.

3. Penumbra's frequent batch auction is the right matching cadence model (Q1) — uniform clearing price per batch kills intra-batch ordering MEV. But its sealed-bid privacy is not shipped, so do not cite Penumbra as proof that encrypted batch auctions work in production. Batch cadence + a working privacy mechanism (MPC or threshold encryption) is the combination to design for.

Dead-ends / cautions

• SUAVE is discontinued (repo archived May 2025) — do not build on it or treat Flashbots TEE-ordering as available product.
• TEE-only privacy (Sapphire, Tristero) trades cryptographic guarantees for hardware-vendor trust and a side-channel attack surface. Sapphire is a usable confidential substrate but ships no matching venue; Tristero and Singularity are unproven — their "truly dark" claims are not verifiable against a live, audited deployment. Treat TEE as a latency optimisation or fallback, never as the sole privacy guarantee for a decentralized engine.
• CoW / UniswapX / 1inch are not dark pools. They are excellent MEV-resistant settlement and batch-auction references, but pre-trade interest is fully visible to the solver/filler set. Reusing their architecture without an added privacy layer would reproduce the operator-visibility problem the framing doc rules out.
• Encrypted mempool (Shutter) is a front-running shield primitive, not a venue, and leaks metadata (gas, address, size). It can harden a transport layer but cannot by itself deliver a persistent dark order book.
• ZK-shielded substrates (Aztec, Starknet, Railgun, Panther) shield balances and transfers, not matching. None provides a matching engine; they are hosting environments, not designs to copy.

Net recommendation for the design phase: the strongest viable foundation is a Renegade-style MPC + collaborative-zkSNARK dark matching core (cryptographic Q2, midpoint-peg Q1/Q5, atomic-match Q3 primitive) + an intent/solver cross- chain settlement layer (UniswapX/Fusion+-style, the proven Q3 cross-chain mechanism). This is the only combination where every component has shipped in production somewhere — even though no single system combines them.

Open questions

1. Cross-chain dark matching has no production precedent. Renegade is dark but single-chain; UniswapX/Fusion+ are cross-chain but not dark. Does running the MPC match across relayers that then trigger two independent chain settlements preserve atomicity, and what is the failure mode if one leg reverts? (Feeds Q3.)
2. Reference price cross-chain (Q5). Renegade pegs to one CEX (Binance) on one venue. For a multi-asset cross-chain cross, what reference is fair and manipulation-resistant when the two legs price on different markets? Workstream on reference-price methodology must resolve this.
3. MPC counterparty discovery vs darkness. Renegade's pairwise MPC requires a relayer to attempt a match to learn there is one — does the handshake-timing metadata leak resting interest? Quantify the leakage.
4. Batch vs continuous for a dark cross-chain venue. Penumbra-style batching bounds MEV and eases cross-chain synchronisation (all legs settle on a batch boundary) but adds latency. Renegade-style continuous crossing is lower latency but harder to synchronise across chains. Which does the framing doc's Q1 favour given cross-chain settlement constraints?
5. Asset-type coverage of atomic settlement. Fusion+ HTLC escrow works for lockable fungible assets; does it extend to the multi-asset scope the framing doc requires, or is solver-capital intent settlement mandatory for the general case?
6. Is FHE-based encrypted matching (Singularity's claim) viable at production latency, or is MPC strictly the practical choice for matching over hidden orders today? No live evidence either way was found.

Sources

1. Renegade Whitepaper (Protocol Specification v0.6), Bender & Kraut — https://whitepaper.renegade.fi/
2. Renegade Docs — The MPC-ZKP Architecture — https://docs.renegade.fi/core-concepts/mpc-zkp
3. Renegade Help — What is a collaborative zkSNARK? — https://help.renegade.fi/hc/en-us/articles/32529961385363-What-is-a-collaborative-zkSNARK
4. Arbitrum Blog — Beyond EVM: Stylus powers Renegade's Onchain ZK Dark Pools — https://blog.arbitrum.io/renegade-stylus-case-study/
5. The Block — On-chain dark pool Renegade goes live on Arbitrum — https://www.theblock.co/post/314959/on-chain-dark-pool-renegade-goes-live-on-arbitrum
6. Renegade Substack — Renegade Goes Live on Arbitrum One — https://substack.renegade.fi/p/renegade-goes-live-on-arbitrum-one
7. Penumbra Protocol Spec — Decentralized Exchange (ZSwap) — https://protocol.penumbra.zone/main/dex.html
8. Penumbra Protocol Spec — Homomorphic Threshold Encryption (flow encryption) — https://protocol.penumbra.zone/main/crypto/flow-encryption/threshold-encryption.html
9. Penumbra Protocol Spec — overview — https://protocol.penumbra.zone/main/index.html
10. CoW Protocol — Understanding Batch Auctions / Fair Combinatorial Batch Auction — https://cow.fi/learn/understanding-batch-auctions , https://docs.cow.fi/cow-protocol/concepts/introduction/fair-combinatorial-auction
11. CoW DAO: 2025 in Review — https://cow.fi/learn/cow-dao-2025-in-review
12. Uniswap Blog — Introducing the UniswapX Protocol / How Dutch Auctions Deliver Better Swaps — https://blog.uniswap.org/uniswapx-protocol , https://blog.uniswap.org/how-dutch-auctions-deliver-better-swaps
13. Uniswap Docs — UniswapX Overview — https://docs.uniswap.org/contracts/uniswapx/overview
14. Uniswap cross-chain aggregator review — https://stablecoininsider.org/uniswap-cross-chain-aggregator/
15. 1inch — A deep dive into 1inch Fusion / Fusion+ cross-chain — https://blog.1inch.com/a-deep-dive-into-1inch-fusion/ , https://blog.1inch.com/1inch-introduces-fusion-plus/
16. Gnosis — Shutterized Gnosis Chain is Live — https://www.gnosis.io/blog/shutterized-gnosis-chain-is-live
17. Shutter Network — Breaking Encrypted Mempool Limitations / Debunking Myths — https://blog.shutter.network/breaking-encrypted-mempool-limitations-with-advanced-cryptography/
18. Mempool Privacy via Batched Threshold Encryption: Attacks and Defenses (eprint 2024/669) — https://eprint.iacr.org/2024/669.pdf
19. Flashbots suave-geth repository (archived 12 May 2025) — https://github.com/flashbots/suave-geth
20. Oasis — Sapphire confidential EVM ParaTime — https://oasis.net/sapphire , https://oasis.net/blog/get-to-know-sapphire-the-industry-first-confidential-evm-paratime
21. Axios — Crypto startup Tristero sets out to build autonomous dark pools — https://www.axios.com/2023/05/02/tristero-autonomous-dark-pools
22. Flashbots Collective — The Crypto Dark Pool Design Playbook — https://collective.flashbots.net/t/the-crypto-dark-pool-design-playbook/3752
23. Cross-Chain Sealed-Bid Auctions Using Confidential Compute Blockchains (arXiv 2510.19491, 2025) — https://arxiv.org/pdf/2510.19491
24. a16z crypto — A cross-chain sealed-bid auction using Aztec Connect — https://a16zcrypto.com/posts/article/cross-chain-sealed-bid-auction-with-aztec-connect/
25. Railgun — On-chain ZK Privacy Ecosystem — https://www.railgun.org/ , https://docs.railgun.org/wiki
26. Panther Protocol — Monthly Update January 2025 / Limited Mainnet Beta — https://blog.pantherprotocol.io/monthly-update-january-2025/
27. Singularity — Why DeFi Needs Darkpools / DarkSwap — https://www.thesingularity.network/post/why-defi-needs-darkpools , https://darkswap.thesingularity.network/
28. Starknet — strkBTC private bitcoin / confidential DeFi — https://www.theblock.co/post/391420/starknet-introduces-strkbtc-to-bring-private-bitcoin-and-confidential-defi-transactions-to-its-layer-2-network