Hardware Security Module (HSM)
A hardware security module (HSM) is a tamper-resistant physical device that generates, stores and uses cryptographic keys inside a certified boundary. Keys created in an HSM can be marked non-extractable, so applications ask the module to sign and never handle the key itself. Banks use HSMs to protect payment keys, and a growing number use them to sign digital asset transactions.

How a Hardware Security Module Works
An HSM is a dedicated computer with one job: to perform cryptographic operations on keys that never leave it. It generates keys from its own entropy source, stores them inside a tamper-resistant enclosure and erases them if someone tries to open it.
Applications reach the module through a standard interface, most often PKCS#11. The application sends data to sign, the HSM signs it with a key it holds, and only the signature comes back. A key generated with non-extractable attributes cannot be exported, even by an administrator.
Control over the module is split. Administrators configure it, usually under M-of-N activation, where several key holders must be present. Operator credentials let an application sign and derive keys but not change the configuration. The module writes its own audit log, which can stream to a security monitoring system so any signature without a matching approved request raises an alert.
Most institutions require certified hardware. FIPS 140-3 is the US and Canadian standard for cryptographic modules, and Level 3 adds physical tamper resistance and identity-based authentication. Common Criteria EAL4+ is the equivalent international evaluation.
HSMs for Digital Assets
Blockchains use different signature schemes than traditional banking systems. Bitcoin, Ethereum and EVM chains sign with the secp256k1 curve, while Solana, Sui and Aptos use Ed25519. Secp256k1 is not a FIPS-approved curve, so an HSM that signs Bitcoin or Ethereum transactions runs with non-approved algorithms permitted. That setting widens the algorithms the firmware will execute. It doesn't remove the module's certification, tamper boundary or key non-extractability.
A digital asset HSM typically holds one master seed per vault and derives every wallet and address from it, using BIP32 for secp256k1 and SLIP-10 for Ed25519. The derivation scheme has to be documented, because recovery depends on finding every address again from the same seed.
An HSM signs whatever it receives. It can prove a key is safe, but it can't tell whether a transaction goes to the right address, whether the right people approved it or whether it passed screening. Those controls have to sit in front of the module, which is where most of the operational work in an HSM program goes.
Backup needs planning before the first key exists. Keys can only be copied to modules in the same cloning domain, which is fixed when the HSM is initialized. Losing the hardware without a clone means losing the keys, so institutions usually run a primary pair for high availability and a second pair for disaster recovery.
HSM Deployment Models
Model | Where the HSM runs | How signing requests reach it | Typical use |
|---|---|---|---|
On-premises | The institution's own data center | Over an isolated network segment from an approved application | Operational wallets where regulation or policy requires keys on owned hardware |
Cloud HSM | A dedicated module in a cloud provider's facility, controlled by the institution | Over a private network connection | Faster deployment without buying and racking hardware |
Air-gapped | An isolated environment with no network path | On removable media, in human-attended signing sessions | Cold storage and long-held balances |
Institutions often combine models, for example an air-gapped HSM for cold storage next to an on-premises or cloud HSM for daily payment flows. Moving assets between them is an ordinary onchain transfer.

HSM vs MPC vs Multisig
Question | HSM | MPC | Multisig |
|---|---|---|---|
Where does the key live? | Whole, inside certified hardware | As key shares on separate machines; the full key never exists | As several independent keys, each signing separately |
What protects it? | The physical tamper boundary and certification | Distribution: no single machine holds enough to sign | Distribution, enforced by the blockchain |
Chain support | Any chain on a curve the module supports | Any chain on a supported curve | Only chains with native multisig or smart contract wallets |
Main trade-off | Hardware cost, procurement time and backup planning | No hardware certification, though shares can run inside secure enclaves | Higher fees and visible signer structure onchain |
The models can also share one architecture. An HSM can sign blockchain transactions while a separate key, held as MPC shares, approves sensitive administrative changes such as new approvers or policy updates. Read more about MPC wallets.
How Utila Can Help
Utila connects institution-controlled HSMs to its wallet infrastructure, starting with Thales Luna. Your wallet keys are generated in your HSM, stay non-extractable, and Utila never receives key material.
A Utila connector runs inside your environment, holds the sealed HSM credentials and passes only approved signing requests to the module.
Utila's policy engine, approval quorums and AML screening run before any request reaches the HSM.
Utila builds each transaction, broadcasts it after your HSM signs and records the approvals and policy evaluation behind it.
On-premises, cloud and air-gapped HSM deployments are supported, alone or side by side.
Read how banks connect HSM signing to their digital asset operations.

