# Privacy for banking subnets

The canonical ledger is fully transparent. Banks will not put client flows on that, and SQRL does not ask them to. Privacy for subnets is delivered in three phases, each built on techniques already proven in production financial networks.

> Part of the SQRL documentation, section "Protocol".
> Rendered page: https://docs.sqrl.tech/protocol/security/privacy

## Phase 1: closed subnets

Confidentiality by data distribution, available with the first bank deployments. Subnet members hold full state; outsiders see only Bitcoin-anchored state-root commitments; supervisors receive observer nodes; issuers enforce whitelists at contract level. No new cryptography is needed. Within a subnet, members see each other's flows, which suits a single institution or a tightly scoped consortium.

## Phase 2: confidential amounts

For shared subnets where competing banks coexist, amounts and asset types are blinded with Pedersen commitments. Totals stay verifiable, because the commitments are additively homomorphic and conservation checks still hold, while individual values are hidden from other members. Auditor and supervisor view keys keep full visibility for oversight.

## Phase 3: zero-knowledge compliance proofs

Zero-knowledge proofs let an institution prove compliance facts, such as whitelist membership, solvency or sanctions-screening status, without revealing counterparties. This phase is deliberately staged last, as the tooling reaches the operational standard a bank deployment requires.

## What SQRL rules out

* **Trusted execution environments as the root of trust.** Repeated enclave key extractions show a single hardware break can expose a chain's entire privacy history. Enclaves may harden a deployment; they are never the trust root.
* **Channel-partitioning schemes** of the kind earlier enterprise ledgers built in, which their own industry has since retired.
* **Pooled custody federations and pegs.** SQRL assets bind to Bitcoin UTXOs directly, and state replays deterministically from the chain.
