Security Overview

An honest summary of what is post-quantum-secured today and what is not. For live/pending status see Status & Roadmap. Last reviewed: 2026-10-06.

RougeChain L1 — post-quantum-secured

  • Accounts, transactions, validator block signatures and staking operations use ML-DSA-65 (FIPS 204). Messaging and mail use ML-KEM-768 (FIPS 203).
  • Transactions are signed client-side; private keys do not leave the wallet.
  • In the web wallet, setting a password (min 8 characters) is required to finish creating or importing a wallet, and keys are written to browser storage only encrypted (AES-256-GCM). A wallet an older version stored unencrypted must be secured with a password before it can be used again. See Create a Wallet.
  • Messages and mail are encrypted to each participant's long-term ML-KEM-768 key. There is no forward secrecy: anyone who later obtains that key can decrypt the stored messages and mail it was used for, both received and sent.
  • These algorithms are quantum-resistant under current cryptographic understanding. No system can promise more than that, and implementation bugs remain possible.
  • Shielded transfers are not available. The V1 shielded pool is suspended since block 245: its STARK proof established value balance only, did not bind a withdrawal to a specific note, and its prover has no zero-knowledge mode, so it did not deliver privacy. The mainnet pool was empty. A redesigned pool (V2, Plonky3-based with a hiding mode, launching under a 1,000,000 XRGE cap) is in development and unaudited; nothing may claim audited privacy until an external audit covers it.
  • Messenger and mail are node-hosted, not on-chain. Since node release 1.6.3 a directory entry can be replaced only by the signing key that owns it, an encryption key registered to another wallet is refused, and the unauthenticated legacy read routes return 410 Gone. Nodes before 1.6.3 do not enforce this.

Signatures commit to the network

Mainnet (rougechain-mainnet-1) and testnet (rougechain-devnet-1) accept the same keys, and the web wallet uses one key on both. From node release 1.6.4 and the wallet releases listed in the changelog, every payload a wallet signs for a node — transactions and signed requests (mail, messenger, names, votes) — carries the chain id of the network it is meant for (chainId; chain_id in a rougechain CLI envelope), inside the signed bytes. A signature therefore commits to one network.

  • Node rule (release 1.6.4, no fork). A node refuses a signed payload that names another chain id (CHAIN_ID_MISMATCH), at the API, at mempool admission and when producing. A payload that names no network is still accepted while wallets update; operators turn on REQUIRE_SIGNED_CHAIN_ID to refuse those too (CHAIN_ID_REQUIRED).
  • Wallets take the chain id from their network setting, compare it once per session with the node's GET /api/health → chain_id, and refuse to sign if the two differ. The extension and Qwalla show the network a request is for and refuse one for a network other than the selected one.
  • Consensus rule (CHAIN_ID_BINDING, not scheduled). From an activation height to be announced, a block is invalid if any transaction's signed bytes do not commit to the chain id; node-signed transactions switch to a network-bound signing format at the same height. Until then consensus is unchanged. See the upgrade schedule.
  • Sign-in messages already carry a Chain ID line; verifiers must check it (Wallet Authentication).
  • Contracts that verify signatures themselves can read the chain id with host_get_chain_id from the CONTRACT_CHAIN_ID upgrade (not scheduled) — see Smart Contracts.

Bridges — mixed

PathToday
RougeChain side of every bridgeML-DSA-65-signed withdrawals
XRGE on Base (BridgeVaultV2)Classical Base-side authorization (Safe multisig + capped relayer key). Hardened R1 architecture.
qETH / qUSDC on Base (RougeBridge)Classical Base-side authorization: single operator owner/release key, with a 2-of-3 Safe multisig guardian that can pause the bridge and cancel queued large releases
Bitcoin (qBTC)Separate bridge; Bitcoin custody uses Bitcoin's classical signatures

So: assets held on RougeChain are protected by post-quantum signatures. The Base-side custody of bridged assets is not post-quantum today. The bridges are operator-run (deposit watcher, relayers and the bridge_mint key), the ETH/USDC contract is owned by a single hot key, and no bridge component has been externally audited. ETH/USDC deposits go through the contract's depositETH / depositERC20 and are claimed automatically; the manual claim route is closed at the public endpoint. Bitcoin deposits are live; Bitcoin withdrawals are switched off in the site.

What V3 will change (not activated)

The V3 XRGE bridge changes the XRGE authorization path on Base to on-chain-verified ML-DSA-65 signatures from an M-of-N authority. It has been built, frozen as an audit candidate and rehearsed, but is not deployed or activated, and it has not yet been externally audited.

  • qETH and qUSDC remain outside the V3 post-quantum scope.
  • The Bitcoin bridge is separate from the Base bridge architecture and from V3.

Finality

Verified BFT finality (FINALITY_V2) is live since block 150: every block carries a certificate of ML-DSA-65 precommits from more than two thirds of stake for its parent, and nodes reject blocks without one. Stake is still concentrated in few keys — three validators are staked and one holds more than 99.9 % of the stake, so its own vote is the quorum — which limits what any BFT guarantee means until the validator set broadens.

No slashing is active. Missed-block slashing is frozen since block 100, the legacy slash transaction is rejected since block 245, and evidence-based slashing does not exist yet. The 10,000 XRGE minimum stake is enforced by the node API, not by consensus.

Decentralization

Validator stake and bridge operation are currently concentrated in a small operator set. This is a known limitation and part of the roadmap: a consensus redesign (rotating proposers, a capped validator set, a consensus-enforced minimum stake, evidence-based slashing) was decided on 2026-10-06 and is not built; see Status & Roadmap.