Aurigraph Research Team · Interoperability · Frontier Tech · status as of 2026-10-20
Tokenized assets are only as useful as the markets they can reach. An asset minted on one ledger but stranded there — unable to settle, collateralize, or trade against liquidity that lives elsewhere — captures a fraction of its potential value. Cross-chain interoperability is the connective tissue that lets value move between ledgers, and it has historically been the single most dangerous surface in the entire crypto stack: bridges have lost more user funds than any other component, repeatedly, at nine-figure scale. Aurigraph's approach treats interoperability as a compliance-preserving, proof-carrying problem rather than a raw fund-transfer problem. Assets originate on Aurigraph — where identity, compliance, and provenance authority permanently reside — and reach external chains through a lock-and-mint rail whose every step is gated by a validator quorum and backed by signed Merkle-inclusion attestations exported from the Aurigraph ledger. Signatures are dual-form: ML-DSA-87 (NIST Category 5 post-quantum) as the canonical Aurigraph signature, plus secp256k1 ECDSA for EVM ecrecover compatibility. This paper is candid about what is implemented versus designed, and about where trust assumptions remain — because in bridging, the honest description of the trust model is the security analysis.
ecrecover
Blockchains are islands. Each maintains its own state, its own consensus, and its own view of who owns what — and, by construction, no chain can natively read another's ledger. That isolation produces two structural problems.
Liquidity fragmentation. Capital pools independently on each chain. A tokenized treasury bill on one ledger cannot be posted as collateral in a lending market on another; a stablecoin deep on one chain is thin on the next. For a general-purpose asset this is an inconvenience; for a regulated real-world asset (RWA) it is close to fatal, because the venues that make a tokenized asset economically useful — deep DEX and lending liquidity — overwhelmingly live on public EVM and Solana ecosystems, not on any single permissioned ledger. An asset that cannot reach that liquidity is a compliance artifact without a market.
Portable assets. Ownership should be able to travel with the holder. A holder who moves from one venue to another, or who wants to use an asset as collateral where the best rate lives, needs the asset — and everything that makes it trustworthy — to move with them. For a plain fungible token, "portability" means moving units. For an Aurigraph asset, it means moving units together with their identity, compliance, and provenance envelope — otherwise a regulated security silently degrades into an unregulated bearer token the moment it crosses a bridge.
Aurigraph's interoperability thesis follows directly: originate compliant here; settle on public liquidity there — without shedding the compliance envelope in transit. Capture the origination and compliance layer; bridge to liquidity rather than trying to rebuild it.
Bridges are the most-attacked component in the industry, and it is worth being blunt about why. The published record is grim: the Ronin bridge lost roughly $625M to compromised validator keys; other cross-chain bridges have suffered comparable eight- and nine-figure losses. The aggregate exposure runs to tens of billions of dollars in cross-chain assets secured by trust models that, on inspection, were far weaker than users assumed.
The failure modes recur:
The design consequence is that a bridge must minimize the number of parties it trusts, verify proofs rather than accept assertions, wait for finality before acting, and — for regulated assets — refuse to let compliance controls fall away at the border. The remainder of this paper describes how Aurigraph's design targets each of these.
Aurigraph's cross-chain rail is built around a small number of load-bearing ideas, most of which exist as primitives in the V12 codebase today.
Lock-and-mint, canonical-on-Aurigraph. The canonical asset and the registry of record stay on Aurigraph, locked in bridge escrow; the external chain receives a representation, not the original. This is deliberate: for a compliance-first origination platform that must retain freeze, revocation, Travel-Rule, and forced-transfer authority over an asset for its entire life — including long-dated instruments — the compliance authority and the asset-of-record must permanently reside in one place. Burn-and-mint (destroy on source, recreate on destination) is appropriate for fungible bridge tokens with no off-chain authority; it is the wrong topology for a regulated RWA. Redemption is the clean inverse: burn the external representation, and the canonical units unlock on Aurigraph, under the invariant that the sum of external representations never exceeds the locked canonical balance.
A guarded state machine. A bridge request advances through explicit, guarded states — INITIATED → LOCKED → RELAYED → CONFIRMED, with terminal FAILED / REFUNDED — enforced server-side. Each transition is role-gated: initiation is an enterprise-role operation; the lock, relay, and confirm legs are validator-role operations. A persisted request record carries the token type (NATIVE, ERC20, RWA, DERIVED), the asset id, the initiator's DID or wallet, the source-chain lock transaction hash, the relay transaction hash, and a SHA-256 evidenceHash over the proof payload as tamper-evident evidence.
INITIATED → LOCKED → RELAYED → CONFIRMED
FAILED
REFUNDED
NATIVE
ERC20
RWA
DERIVED
evidenceHash
Signed Merkle-inclusion attestations. Rather than trusting a relayer's word that state exists on Aurigraph, the destination side verifies a Merkle inclusion proof against the Aurigraph identity/compliance ledger root. Aurigraph's CrossChainProofBridge resolves the relevant record, fetches its Merkle inclusion proof, and exports it in chain-native form: for EVM, an ABI-encoded bytes32[] proof with merkleRoot and leafHash; for Solana, a Borsh-style proof. Each export carries two signatures — an ML-DSA-87 post-quantum signature as the canonical Aurigraph attestation, and a secp256k1 ECDSA signature that a Solidity ecrecover can verify on-chain. Stated plainly: in the current CrossChainProofBridge the ECDSA attestation is produced with a fresh ephemeral key pair per export rather than a persisted, published attestation key, so an external verifier cannot yet pin a stable Aurigraph signer (AV11-3853); replacing it with a persisted ML-DSA attestation key is tracked work.
CrossChainProofBridge
bytes32[]
merkleRoot
leafHash
Validator quorum on the relay. No single relayer or oracle can authorize a mint or a freeze. The relay leg requires a validator quorum — an EIP-712-signed, m-of-n threshold in which unique validator public keys are counted, so duplicate signatures cannot inflate a count. In the current cross-chain engine that threshold is 6-of-9 (QuorumValidator.DEFAULT_THRESHOLD); note that the QuorumValidator class itself only counts well-formed signatures and the real ECDSA verification is performed in CrossChainBridgeServiceImpl. The productized bridge targets a 67% validator quorum (a design target, not yet delivered), aligned with the platform's consensus posture. Either way, the guarantee is the same in spirit: crossing the honest-majority threshold is required to move value, and a minority of malicious validators cannot rob the bridge.
QuorumValidator.DEFAULT_THRESHOLD
QuorumValidator
CrossChainBridgeServiceImpl
Persisted relay proofs. Every relay produces a durable BridgeRelayProof record before the operation returns success. This is treated as financial-loss-critical: a lost relay proof is a lost audit trail for a value movement, so persistence is a hard precondition of the relay leg, not a side effect.
BridgeRelayProof
Reorg and replay guards. Bridge messages carry a nonce as a replay guard and track a finality depth against an observed depth; a source event that has not reached required finality parks in a REORG_PENDING state rather than triggering a premature mint. This directly targets the reorg-double-mint failure mode.
REORG_PENDING
Design direction (labelled, not yet shipped as light-client verification). The strongest form of trust minimization is a light-client / SPV-style proof, in which the destination verifies source-chain block headers directly rather than trusting a relayer's attestation. The V12 tree includes early scaffolding toward header/template verification, but today's rail is grounded on the signed Merkle-attestation + validator-quorum model described above, not on full on-chain light-client verification. We call this out explicitly so the trust model is not overstated: the current mechanism trusts the validator quorum's signatures over a Merkle proof, not an independent header chain. Moving from quorum-attested proofs toward light-client verification is a roadmap direction (§7), not a present guarantee.
What makes Aurigraph's interoperability distinctive is not that units move — every bridge moves units — but that the compliance envelope moves with them. Aurigraph assets are minted as Composite Tokens: a single W3C Verifiable Credential binding decentralized identity, regulatory compliance, provenance, a cryptographic binding to the underlying artefact, and post-quantum key material into one envelope (see the Composite Token Framework).
When such an asset bridges, the destination representation is deliberately not a bare token. On EVM it is minted as an ERC-3643 (T-REX) permissioned token; on Solana as an SPL Token-2022 with a transfer-hook. Both enforce, on every transfer, that the recipient is a verified holder. The Aurigraph-side identity and compliance state — the allowlist, claim topics and trusted issuers, jurisdiction restrictions, holder caps, freeze/pause flags, and Travel-Rule proofs — remains the source of truth, and the external-chain artifacts are projections kept in sync by an attestation oracle. The same CrossChainProofBridge that proves a lock also proves who the holder is and what they are cleared for, so the destination mint can confirm the recipient is on the allowlist before minting.
The consequence for the threat model is decisive: compliance leakage is closed by construction. The bridge never emits a bare ERC-20 or bare SPL token; settlement is restricted to compliance-honoring venues (permissioned or ERC-3643-aware DEXs, hook-honoring Solana RWA venues). Generic permissionless AMM pools — which cannot enforce the allowlist — are out of scope by design, because reaching them would require a bare wrapper that strips the envelope. Aurigraph does not promise the asset appears in every pool; it promises the asset reaches public liquidity without losing its regulatory status.
Honesty about trust assumptions is the whole game in bridging. Here is what the design does and does not assume.
What is trusted. The current rail trusts (a) the validator quorum — a mint or freeze requires the m-of-n threshold of EIP-712 validator signatures, so no single party can forge one; and (b) the attestation oracle's liveness — that it pushes fresh compliance state, including freeze and revocation deltas, to the external chains promptly. It does not trust a single relayer, a single custodian key, or an unverified destination assertion — proofs are Merkle-verified, and value movements are quorum-signed.
The hard problem: revocation/finality lag. The dominant residual risk is the propagation window. When Aurigraph freezes a holder or revokes their KYC, the external-chain representation can still move during the interval before the freeze delta propagates. The mitigations are layered: low-latency freeze/revoke pushes; a fail-closed posture in which the external side auto-pauses transfers if it has not seen a fresh ledger root within a staleness bound; an on-chain pause() kill-switch; and — critically — the lock-and-mint topology itself, which means Aurigraph can always freeze the canonical units immediately, regardless of external-chain state.
pause()
Other explicit limits. Escrow of the locked canonical units is a custody surface, mitigated by permissioned escrow under governance multisig, PQC signing, and a DMRV audit trail. The supply invariant (external representations ≤ locked canonical) must be enforced by the quorum and redemption accounting to prevent inflation. Cross-jurisdiction conflicts (a holder verified on Aurigraph but restricted at the destination venue) are handled by projecting jurisdiction rules to both sides. And a Travel-Rule gap on the public side is designed to be closed by requiring a zero-knowledge Travel-Rule proof reference for transfers above threshold — a design item, because the platform has no real zero-knowledge prover today (its ZK engines fail closed rather than emit a fake proof).
Post-quantum posture. The canonical attestation signature is ML-DSA-87 (NIST FIPS 204, Category 5) — chosen because a ledger recording title to a 30-year asset cannot rest on ECDSA alone, which is breakable by a sufficiently capable quantum computer. The secp256k1 ECDSA signature exists only for EVM ecrecover compatibility; the ML-DSA-87 signature is the durable one. This is a genuine cryptographic guarantee on the attestation path, distinct from the operational trust in the validator quorum's honesty, and we keep the two separate on purpose.
Implementation status — read this plainly. The bridge rail — the request entity, the guarded state machine, relay-proof persistence, the quorum validator, and the CrossChainProofBridge attestation exporter — exists in the V12 tree, with real cryptography (BouncyCastle) on the security-validation path. The transport adapters are explicitly fakes in that tree (CrossChainBridgeFakeEngine, LayerZeroFakeAdapter, CcipFakeAdapter): they model the message flow and never touch an external chain. The RWA public-liquidity bridge that composes these into full lock-and-mint settlement on external chains is at DESIGN status (tracked under epic AV11-3055): the on-chain Solana transfer-hook program, the T-REX deployment on a live EVM, verifier contracts on any public network, and the attestation-oracle wiring are phased work, not shipped mainnet features. Nothing in this paper should be read as a claim that regulated assets round-trip to public-chain mainnet liquidity today.
CrossChainBridgeFakeEngine
LayerZeroFakeAdapter
CcipFakeAdapter
The productized bridge names the following chains. Consistent with §5, treat "active" here as the product target for the settlement rail; the compliance-preserving public-liquidity path across these chains is phased design work.
Portable asset classes span ERC-20 tokens, ERC-721 NFTs, and — the distinctive case — Aurigraph-originated RWAs carried as compliance-preserving ERC-3643 / Token-2022 representations. The intended production transports are LayerZero v2 and Chainlink CCIP, modelled today behind a bridge-adapter interface.
Interoperability is where tokenized value either becomes useful or stays stranded — and, historically, where it most often gets stolen. Aurigraph's answer is to refuse the false choice between reach and safety. By keeping the canonical asset and its compliance authority permanently on Aurigraph, moving value through a lock-and-mint rail whose every step is quorum-gated and backed by post-quantum-signed Merkle proofs, and minting only compliance-preserving representations on external chains, the design targets each of the failure modes that have defined the bridge-hack era: it minimizes trusted parties, verifies proofs instead of accepting assertions, waits for finality, and never lets the compliance envelope fall away at the border. Equally important is the discipline of saying clearly what is built (the rail and its cryptographic primitives) versus what is designed (full public-liquidity settlement, and the move toward light-client verification). That honesty about the trust model is not a caveat bolted onto the security analysis — for a bridge, it is the security analysis.
For the ledger, consensus, and cryptographic substrate this rail is built on, see the Aurigraph DLT whitepaper.
docs/design/DESIGN_RWA_PUBLIC_LIQUIDITY_BRIDGE_AV11-3055.md
aurigraph-v12/src/main/java/io/aurigraph/v11/bridge/
CrossChainBridgeRequestEntity
This document is a technical whitepaper. Claims about the bridge rail, its state machine, quorum, relay-proof persistence, and the CrossChainProofBridge attestation exporter are grounded in the V12 codebase and platform design docs. Full public-liquidity settlement on external-chain mainnets, light-client verification, and live LayerZero/CCIP transports are labelled design/roadmap items — not shipped guarantees. Any figure not backed by a published benchmark should be read as a capability target.
Explore: About · Products · Features · Whitepapers · Sustainability · Tokenized carbon credits · Real-world asset tokenization · SDK · Blog · Plain-text site index · Sitemap · Contact: contact@aurigraph.io