Privium is the omnibus account for Solana traders. Traders deposit into one shared Pool, hold their balances in an encrypted Ledger, and submit encrypted Intents that a trusted-execution Coprocessor crosses against each other at the oracle mid every batch. Only the net Residual of each pair reaches public liquidity, as one Jupiter swap from the Pool, and each batch ends with a Commit that the Solana program accepts after checking the enclave’s signature and the conservation of every asset. This chapter states the problem Privium solves, the structure it borrows from equity brokerage, the two recent pieces of work that make that structure buildable on Solana, the market it serves, and what the rest of this paper contributes.
Every trader is a feed
A Solana wallet is a public, append-only record of everything its owner has done with it. Every swap names the signer, the input and output mints, the amounts and the slot. Indexers decode these records within moments of confirmation, and a whole category of software is built on reading them back. The terminals that dominate Solana spot trading sell wallet tracking as a headline feature: Axiom offers unlimited tracked wallets with per-wallet PnL, win rate, trade history and real-time buy and sell alerts, and its copy-trading engine mirrors a tracked wallet’s buys with fixed-SOL or percentage sizing [5]. A trader with a good record therefore publishes, with every fill, a signal that other traders’ software is configured to act on.
The consequences compound. A wallet that is followed is front-run by its followers on entry and dumped on by them at exit, because the followers’ bots react to the same on-chain event at machine speed and compete for the same thin liquidity. A wallet that is merely visible leaks its size, its holding period and its preferred venues to anyone who wants to price against it. And a swap that is visible before it lands can be sandwiched: a measurement study of Jito bundles over four months of early 2025 found 521,903 sandwich attacks costing victims more than $7.7M, and found users spending more than $2.4M on defensive bundling to avoid them [8]. Those figures are lower bounds by the authors’ own method, since they count only canonical three-transaction bundles and only SOL-denominated legs [8].
Traders already pay to escape this. The standard practice among active Solana traders is to rotate burner wallets: fund a fresh address, trade from it until it becomes known, abandon it, repeat. Rotation is a manual, lossy imitation of a property that equity markets give every client by default. It costs transfer fees and attention, it fragments a trader’s history across addresses they then cannot prove they own, and it fails the moment the funding path is traced, because each fresh wallet is funded from somewhere.
We can state the problem precisely. Let a trader submit a sequence of trades . On a public chain the observable trace of is
one record per trade, each carrying the same address . Any function of a trader’s strategy that can be estimated from (1.1) is public. The aim of Privium is to replace (1.1) with a trace in which no record carries a trader at all. Under Privium the observer sees, per Crossing and per pair ,
that is, the new Ledger root, one net Residual per pair, and the vault balances that follow from deposits, Exits and routed Residuals. No component of (1.2) is indexed by . Deposits and Exits remain visible events at the boundary of the Pool, and Chapter 8 measures exactly how much they reveal and how Privium reduces it.
The omnibus account
Equity markets met this problem long ago and settled it with an account structure. In an omnibus account, a broker holds the cash and securities of many clients in a single account at the clearing firm and keeps the per-client records on its own books [7]. US broker-dealer rules treat such an account as one customer of the carrying firm, while the per-client records stay with the broker that maintains it [14]. The market sees the broker. It does not see which client bought, how much, or when.
The omnibus structure carries a second mechanism that matters as much as the first. A broker that holds many clients’ order flow can cross opposing orders internally, at or inside the best quote, before anything reaches an exchange, and route only the remainder outward. In 2025, trading reported to the US Trade Reporting Facilities, that is, trading away from the lit exchanges, rose 361 basis points to 50.6% of total consolidated US equity volume, the first year off-exchange trading exceeded half the market [6]. More than half of the volume in the deepest equity market in the world now executes away from the exchanges, much of it against the internal books of the firms that hold the orders.
The structure won for two reasons that carry directly to Solana. Clients receive price improvement, because a crossed fill at the midpoint pays neither the spread nor the price impact of walking a book. Clients receive information protection, because the market sees aggregated, netted flow rather than each client’s intent. The omnibus account delivers both from the same fact: many clients’ orders meet inside one account before any of them meets the market.
Three things kept this structure off public chains. First, a broker’s books are private because they are held by a firm, and a public chain has no private books. Second, a broker’s client claims are enforced by law and by the custodian; a protocol needs a claim that is enforced by code even if its operator disappears. Third, internal crossing needs a computation over every client’s private order, and that computation must be both hidden from the public and verifiable by it. Each of these was an open engineering problem on Solana until recently. The next section describes the two pieces of work that closed them.
The thesis from two sources
Private balances as a native primitive
Helius’s “Solana Privacy Rings” [1] argues that public-ledger transparency blocks ordinary financial use of Solana, and that privacy is the chain’s next frontier. It introduces Rings: programmable shielded pools whose balances are encrypted on chain and whose transactions execute natively on Solana and are verified by zero-knowledge proofs [1]. A permissionless default Ring is live on mainnet, encrypting assets and amounts while wallets remain visible, with balances readable only by the owner through a viewing key [1]. Rings accept SOL, SPL tokens and Token-2022 tokens, keep the owner’s signature as a requirement for every transaction, and store their encrypted data on Solana’s own ledger [1].
The article also specifies Custom Rings, aimed at institutions and targeted for mainnet in Q4 2026 [1]. A Custom Ring can be configured as confidential (counterparties visible, balances encrypted) or anonymous (both encrypted), can grant an auditor key read access under a configured policy, and lets a user disclose all of their activity through a viewing key or a single transaction through a transaction viewing key [1]. Helius lists Jupiter among the integrations in progress [1].
For Privium, Rings supply the boundary of the Pool. A trader can fund a Privium deposit from a Ring and take an Exit into one, so that neither edge of a trader’s activity needs to touch a public balance. Rings also fix the vocabulary of selective disclosure on Solana: the viewing key and the per-transaction disclosure key are already familiar objects to wallets and compliance providers, and Privium’s Viewing keys and Receipts (Chapter 11) are designed to sit beside them.
The same period restored the lower layer that Solana’s confidential token tooling rests on. Token-2022 confidential transfers, which hide balances and transfer amounts using the ZK ElGamal proof program, were disabled on mainnet in June 2025 after researchers found a flaw in that program’s Fiat-Shamir transformation; the fix shipped in Agave v2.1.21 in March 2026, and the feature gate was reopened on 17 June 2026 [9]. Chapter 3 treats the twisted ElGamal and sigma-protocol machinery of that stack in full.
Programs over private balances at trading speed
A private balance is not yet a private trading account. Crossing two traders’ orders requires reading both orders, comparing them with a price, and writing both balances, all without revealing any of it. Lokhande et al.’s Encifher [2] is a confidentiality coprocessor for Solana that makes this kind of computation practical. It takes the ciphertext-handle and symbolic-execution model of homomorphic coprocessors such as Zama’s fhEVM and runs it inside a trusted execution environment instead of a homomorphic evaluator [2]. On-chain programs manipulate only 128-bit ciphertext handles; an off-chain coprocessor in a hardware enclave fetches the ciphertexts from a public data availability layer, decrypts and computes on them inside the enclave, re-encrypts the results, and posts attested, Merkle-anchored outputs back on chain [2].
Two features of Encifher’s design are central to Privium. The first is that Solana’s account model already supplies the dependency graph of the confidential computation: because every transaction declares its read and write sets, the Sealevel scheduler orders encrypted operations correctly and in parallel without a custom scheduler [2]. The second is that no single enclave holds the only copy of the decryption capability: trust is spread across a threshold-decryption committee, and enclave attestation is checked on chain [2]. Encifher states its trust base plainly as TEE integrity, honest threshold key management and the cloud attestation root [2]. Privium adopts that trust base and states each of its guarantees against it (Chapter 2).
The reported production figures place this approach in the cost range of ordinary trading. A confidential swap settles in one Solana transaction at 157,000 compute units; the AES-256-GCM encryption layer exceeds one million operations per second; threshold decryption takes single-digit to tens of milliseconds; and the system has served more than 5,000 users and 50,000 confidential operations in production, across confidential payments, swaps and a cross-chain bridge [2].
What neither source builds
Rings make a private balance a native object. Encifher makes a program over private balances fast enough for trading. Neither carries the two to the place where a trader’s information actually leaks, which is the trading account itself. A Ring hides what a wallet holds, but a trade from that wallet still lands as that wallet’s swap, and an Encifher swap is a confidential swap between parties rather than a shared account in which many traders’ flow is netted before it meets the market.
Privium composes them into the omnibus account. The Pool is one program-owned account per asset. The Ledger is Encifher-style encrypted state, one ciphertext leaf per trader and asset, anchored by a Merkle root. The Coprocessor is an attested enclave that does what a broker’s internalization desk does: it decrypts the batch’s Intents, crosses opposing flow at the oracle mid, computes one Residual per pair, and routes it. The Council plays the role that law and the custodian play for a broker’s client: it holds threshold shares of the decryption key and runs the Escape if the Coprocessor stops, so that every trader’s claim to their last committed balance is enforced without the operator. Rings sit at the boundary, so that deposits and Exits can be unlinked as well.
Why now
The structure needs five platform capabilities at once. Each is now available on Solana mainnet, and the first two arrived in 2026.
| Capability Privium needs | What provides it | Status |
|---|---|---|
| Encrypted balances and private deposits and exits | Helius Rings; Token-2022 confidential transfers on the ZK ElGamal proof program | Default Ring live on mainnet; Custom Rings targeted for Q4 2026 [1]; confidential transfers re-enabled 17 June 2026 [9] |
| Computation over encrypted state at trading cost | TEE coprocessor with ciphertext handles and on-chain attestation checks | In production, 157,000 CU per confidential swap [2] |
| Correct ordering of confidential state transitions | Sealevel’s declared read and write sets | Used as the confidential dependency graph [2] |
| One swap from the Pool across all Solana liquidity | Jupiter Swap API V2, whose Router path /swap/v2/build returns instructions for custom transactions |
Current API generation [15] |
| Enclave signatures checked by the program | Solana’s native signature verification programs, read by instruction introspection | Chapter 5 and Chapter 9 |
The table also explains why the structure could not have been built earlier from any single source. Without encrypted balances, an omnibus Pool on a public chain is only a commingled account whose per-trader state must live somewhere off chain and in the clear. Without an attested coprocessor, internal crossing requires either a trusted operator who sees every order or a cryptographic protocol (fully homomorphic encryption or secure multi-party computation) whose latency Encifher’s authors judge too high for interactive use [2]. Without a liquidity aggregator that composes into a program’s own transaction, the Residual could not be executed from the Pool’s vaults as one atomic, enclave-bounded step whose fill the Commit then absorbs. Chapter 9 gives the instruction-level construction of that step.
Market evidence
The venue
Solana is the largest spot venue in crypto by DEX volume. Its decentralized exchanges cleared $197B of spot volume in Q3 2026, up 16.6% from Q2, leading every other chain for the fifth consecutive quarter [3]. In one week of September 2026 Solana processed about 208M DEX trades, against about 190M trades on the NYSE in the same week, a comparison of trade counts rather than dollar value [3]. DefiLlama’s DEX dataset records $49.8B of Solana DEX volume over the 30 days to 15 November 2026 and $969.5B over the trailing twelve months [10].
The trade count matters as much as the dollar figure. Internal crossing produces value in proportion to how often opposing flow arrives in the same batch, and that frequency grows with the number of independent orders per unit of time. Chapter 7 defines the crossing rate, and Chapter 12 derives it as a function of order flow and turns it into an expected-cost comparison.
The terminals
The flow is concentrated in trading terminals, and those terminals are large businesses. DefiLlama records about $18.4M of fees and $8.4M of revenue for Axiom over the 30 days to 15 November 2026, and $475.8M of cumulative revenue [4]. Photon reports more than $40B of cumulative volume and 1.3 million users [5]. Axiom’s base trading fee is 0.95%, falling to 0.75% at its top tier, and Photon’s is 1% flat [5]. These fees are paid on top of the DEX fees and price impact of each fill, which is the cost Privium’s internal crossing removes for the matched portion of every batch.
The same terminals are where the wallet-tracking problem is sold as a product. Wallet tracking with PnL and alerts, and copy trading of tracked wallets, are standard features of the leading terminal [5]. Every serious trader on these platforms is simultaneously a user of the tracking tools and a target of them.
The behaviour already exists
The demand for the omnibus account is visible in what traders already do without it. Burner-wallet rotation is the manual form. Defensive bundling, which the IMC ‘25 study measured at more than $2.4M of user spend over four months [8], is the form traders adopt against ordering attacks. MEV-protection settings in mainstream Solana trading apps, offered through Jito, are a third [8]. Each of these is a trader paying to reduce the information content of their own trace (1.1). Privium removes the trader from the trace instead.
The equity precedent gives the scale of the endpoint. When clients are offered an account that internalizes their flow, a majority of volume moves there: off-exchange trading reached 50.6% of US equity volume in 2025 [6]. Privium does not need that share to be valuable; it needs enough concurrent flow per batch for opposing Intents to meet, and Chapter 7 shows that the value of a Crossing to each participant grows with the flow of the others.
Landscape
Batch matching and private trading each have strong precedents, and Privium borrows from them openly.
CoW Protocol collects signed intents off chain and runs batch auctions in which solvers first look for coincidences of wants inside the batch, matching opposing orders peer to peer without touching on-chain liquidity, and source the remainder from on-chain and off-chain liquidity [11]. Privium’s Crossing is the same economic move, with two differences of construction: the batch is computed inside an attested enclave over encrypted Intents rather than by competing solvers over public orders, and settlement is against a shared encrypted Ledger rather than per-order token transfers.
Penumbra batches swaps per trading pair in each block and writes the batch result to public state, with users claiming their pro-rata share of outputs; its sealed-bid batch swap design reveals only the net flow across a pair in each batch [12]. Privium’s observable trace (1.2) has the same shape, one net quantity per pair per batch, delivered on Solana and against Solana’s existing liquidity.
Renegade is an on-chain dark pool that matches orders with multi-party computation, proves settlement with zero-knowledge proofs, and executes crosses at the midpoint [13]. Privium shares the midpoint crossing rule and the commitment to pre-trade and post-trade privacy, and takes the coprocessor route to the computation, which Encifher has shown to run at Solana transaction cost [2].
Encifher itself runs confidential swaps on Solana [2]. Privium is built on its model and occupies a different place in the stack: not a confidential swap between parties, but the account a trader lives in, with one deposit, the whole token universe that Jupiter routes, internal crossing of everyone’s flow, and no wallet to follow.
Contributions of this paper
This paper specifies Privium completely enough to build, audit and price. Its contributions are as follows.
An account structure. We define the omnibus account on Solana: a single Pool program custodying per-asset vaults, an encrypted per-trader Ledger committed as a Merkle root over ciphertext leaves, and a Coprocessor that advances the Ledger in batches. Chapter 2 gives the system and its properties; Chapter 4 gives the Ledger’s state transition function.
A conservation theorem enforced on chain. We state and prove that for every asset and every committed root, the sum of Ledger balances equals the vault balance net of committed Exits not yet released and deposits not yet credited, and we show how the Solana program checks the public side of that identity on every Commit without seeing a single balance. Chapter 4 proves the theorem; Chapter 9 implements the check.
An attested Coprocessor with on-chain key binding. We bind the Coprocessor’s signing key to its enclave measurement through the report data of a remote attestation quote, register it on chain, and verify every Commit against it. Chapter 5 gives the chain of trust, rotation and upgrade by re-attestation.
A threshold Council and a forced Escape. We give a distributed key generation for a twisted ElGamal key, threshold re-encryption with proofs of correct decryption, proactive resharing, and an Escape protocol that pays every owner their last committed balance when no Commit lands within the liveness window , with a safety theorem. Chapter 6 covers all of it.
A crossing engine with fairness and cost results. We give the netting rule per pair, clearing at the oracle mid under limit constraints, allocation rules with proofs of their fairness properties, an expected-cost inequality between crossing and routing, a Residual routing step bounded by a slippage limit the enclave signs, and an analysis of ordering attacks against batched Residuals versus individual swaps. Chapter 7.
An unlinkability analysis with a formal game. We model a chain observer, a wallet tracker and a timing analyst; measure the anonymity set with min-entropy and Shannon entropy as functions of Pool activity; bound deposit-to-exit, amount and timing linkage; and prove the unlinkability theorem Privium achieves, including with Ring Exits. Chapter 8.
A complete Solana implementation design. We give account layouts with sizes, PDAs and seeds, every instruction with its checks, compute budgets, address lookup tables, write-lock design for Sealevel parallelism, signature verification by instruction introspection, and the Jupiter composition for the Residual (Chapter 9), and the Coprocessor service, oracle ingestion, indexer and the TypeScript SDK with its one-line call (Chapter 10).
Selective disclosure without a wallet. We define Viewing keys, per-transaction disclosure, and Receipts: enclave-attested statements of a trader’s realized PnL over a period, verifiable by anyone and shareable without revealing an address. Chapter 11.
An economic model and a token. We price crossed, routed and direct execution as expected costs, derive the crossing rate’s effect on cost, and specify the fee flow: each Commit pays the Crossing’s fees to the token’s launch wallet, and a crank buys $PRV on the market with them and burns what each purchase delivered. Chapter 12.
Notation and conventions
The notation below is shared by every chapter. Chapters that need further symbols define them where they first appear and keep them local.
| Symbol | Meaning |
|---|---|
| , , | trader, asset, quote asset (USDC by default) |
| , | Ledger balance of in after Crossing ; vault balance of |
| , , | Crossing index; batch interval in slots; liveness window in slots |
| Intent: trader, pair , side , size in base units, limit , expiry slot , nonce | |
| , , | oracle mid, clearing price, Residual for pair in Crossing |
| fee rate, 0.10% (10 bps) | |
| , , , | ristretto255 group, its order, generators |
| , | hash function; AEAD (AES-256-GCM) |
| , | Council size and threshold |
| , | Coprocessor signing key pair; Ledger root after Crossing |
| , | security parameter; negligible function |
The letter denotes the hash function in protocol descriptions and Shannon entropy in the unlinkability analysis; context always fixes which. Amounts are integers in each asset’s base units, and prices are quoted in the quote asset per unit of the base asset.
Privium design parameters appear with a recommended value marked “(parameter)” and a one-line reason, in the chapter that owns them. Engineering estimates are marked “est.” and derived where they appear. Every other number is cited.
Reading guide
The paper is written for three readers, and each can take a direct path through it.
| Reader | Path | What it covers |
|---|---|---|
| Investor or product reader | Chapters 1, 2, 7 (first two sections), 12 | the structure, the life of a trade, why crossing is cheaper, fees and $PRV |
| Protocol engineer | Chapters 2, 4, 5, 9, 10 | state, attestation, the Solana program, the Coprocessor and SDK |
| Cryptographer or auditor | Chapters 2, 3, 4, 6, 8, 11 | primitives and reductions, conservation, threshold keys and the Escape, unlinkability, Receipts |
Chapter 2 is the common entry point: it introduces every actor, follows one trade from deposit to Exit, draws the boundaries between trust domains, and states the six properties the rest of the paper proves. Chapters 3 to 6 build the confidential state and the keys that protect it. Chapters 7 and 8 analyse what Privium does with that state, first for price and then for privacy. Chapters 9 and 10 turn the design into code. Chapters 11 and 12 cover what a trader can prove about themselves and how the protocol is paid.
References
- “Solana Privacy Rings”, Helius, 2026. https://www.helius.dev/blog/solana-privacy-rings
- N. Lokhande, R. Gupta, R. Chahar, A. Jangra, A. Prajapati, M. Kumari, “Encifher: A Trusted-Execution Coprocessor for Confidential Computation on Solana”, IACR ePrint 2026/1504, July 2026. https://eprint.iacr.org/2026/1504
- V. Nguyen, “Solana leads all chains in spot DEX volume for fifth straight quarter with $197B in Q3 2026”, Crypto Briefing, 2 October 2026. https://cryptobriefing.com/solana-leads-spot-dex-volume-q3-2026/
- “Axiom: fees and revenue”, DefiLlama (fees and revenue API, data to 15 November 2026), accessed 16 November 2026. https://defillama.com/protocol/axiom
- “Photon vs Axiom: Solana Memecoin Terminal Comparison (2026)”, MadeOnSol, 31 May 2026, updated 30 September 2026. https://madeonsol.com/blog/photon-vs-axiom-solana-trading-terminal
- “2025 U.S. Equities Year in Review”, Cboe Global Markets, 26 January 2026. https://www.cboe.com/insights/posts/2025-u-s-equities-year-in-review
- “Omnibus account”, Nasdaq Glossary, accessed 16 November 2026. https://www.nasdaq.com/glossary/o/omnibus-account
- N. Gerzon, B. Weintraub, J. In, A. Mislove, C. Nita-Rotaru, “Quantifying the Threat of Sandwiching MEV on Jito: A Measurement of Solana’s Leading Validator Client”, Proceedings of the 2025 ACM Internet Measurement Conference (IMC ‘25), October 2025. https://doi.org/10.1145/3730567.3764493
- “Confidential transfers return to Solana mainnet after year-long security pause”, Solana Compass, 18 June 2026. https://solanacompass.com/news/confidential-transfers-return-to-solana-mainnet-after-year-long-security-pause
- “DEX volume: Solana”, DefiLlama (DEX overview API, data to 15 November 2026), accessed 16 November 2026. https://defillama.com/dexs/chain/solana
- “Coincidence of Wants”, CoW Protocol documentation, accessed 16 November 2026. https://docs.cow.fi/overview/coincidence-of-wants
- “Batch Swaps”, Penumbra Protocol Specification, accessed 16 November 2026. https://protocol.penumbra.zone/main/dex/swap.html
- “Dark pool explainer”, Renegade documentation, accessed 16 November 2026. https://docs-v2.renegade.fi/concepts/dark-pool-explainer
- “SEA Rule 15c3-3 and Related Interpretations”, FINRA, accessed 16 November 2026. https://www.finra.org/rules-guidance/guidance/interpretations-financial-operational-rules/sea-rule-15c3-3-and-related-interpretations
- “Swap API”, Jupiter Developer Docs, accessed 16 November 2026. https://developers.jup.ag/docs/swap-api