I built a cross-chain protocol with threshold signatures.
Crossflow is a decentralised cross-chain loan protocol on a custom Cosmos SDK / CometBFT L1. Borrowers lock collateral on Bitcoin, EVM, or Solana; the chain issues loans without a custodial multisig. The signing key never exists in one place. Threshold signatures (Gennaro-Goldfeder 2018 MPC-ECDSA) let a quorum of attesters co-sign outbound transactions without any single node holding the full secret.
Most cross-chain bridges are trusted multisigs: a small set of operators hold keys, sign withdrawals, and collect $600M into one target every 18 months. The fix is not better key management. It is never assembling the key at all. Threshold signatures let a quorum of validators jointly produce a valid ECDSA or EdDSA signature through a multi-party computation ceremony, so the private key provably never exists in one place. Crossflow ships that primitive on a purpose-built Cosmos SDK chain: borrow against cross-chain collateral with no custodian and no bridge operator to bribe.
Cosmos SDK / CometBFT L1 (CFN_testnet-1 · uCFN)
├── x/loan loan lifecycle · collateral accounting
├── x/tss keygen · resharing · signing coordination
├── x/lock per-chain lock records
├── x/asset supported asset registry
├── x/chain external chain config
├── x/oracle IBC price feed consumer
└── x/reward attester incentives
│
├──▶ observer daemon [ go ]
│ monitors Bitcoin / EVM / Solana for lock events
│ posts observation votes to x/loan on L1
│
└──▶ attester daemon [ go ]
receives sign requests from L1
executes Gennaro-Goldfeder MPC-ECDSA ceremony
├── secp256k1 → BTC (HTLC scripts + custom sighash)
│ EVM (loan · repay · supply txs)
└── ed25519 → Solana
│
▼
outbound signed transaction
broadcast to target chain
BandChain ──[ IBC ]──▶ x/oracle → collateral price feeds
A t-of-n multisig requires each signer to hold a complete key share that can be extracted under duress. GG18 MPC-ECDSA never produces a full key. Signers run a distributed ceremony, so no single compromise breaks the scheme. The security baseline was established quickly by adapting the Kudelski-audited tss-lib rather than implementing the paper from scratch.
EVM L2s inherit Ethereum's account model and consensus, which is useful for DeFi composability but wrong for a chain that needs native module governance and IBC. Cosmos SDK lets x/tss and x/loan live as first-class state machines with deterministic execution and direct IBC transport to BandChain, without wrapping logic in Solidity contracts.
Signing (attester) and chain monitoring (observer) have completely different failure modes. Separating them means a stalled observer scan doesn't block an in-flight signing ceremony, and each daemon can be upgraded, restarted, or replicated independently without touching the other.
Bitcoin's UTXO model has no native smart contracts. HTLCs (hash time-locked contracts) encode the lock condition directly in the script with a custom sighash, giving atomic cross-chain settlement guarantees without a bridge operator. If the secret is not revealed within the timelock, funds return to the sender.
A centralised price feed is a single point of manipulation. One compromised endpoint can trigger mass liquidations. BandChain is a decentralised oracle chain with its own validator set. Consuming price data over IBC means price updates arrive as verifiable IBC packets, not as trusted HTTP calls.
Bitcoin and EVM chains sign over secp256k1; Solana signs over ed25519. One attester fleet handles both curves from a single keygen ceremony per curve. Two separate fleets would double the operational surface: upgrades, monitoring, and key ceremonies run twice for no security benefit.
Crossflow reached testnet (CFN_testnet-1) with a working keygen, resharing, and per-transaction signing ceremony across Bitcoin, EVM, and Solana. The engineering demonstrates implementing Gennaro-Goldfeder MPC-ECDSA (Feldman VSS, Paillier encryption, Schnorr/DLN zero-knowledge proofs, MtA sub-protocol) on a custom L1 chain, integrating HTLC scripts on Bitcoin for atomic settlement, and operating a decentralised oracle over IBC. That is the primitive stack that bridge security and institutional DeFi teams spend years hiring for. Built as one engineer at Hexmount.