(Security)

Designed for the day your keys leak.

The threat palmfi designs for is an attacker who can produce valid signatures, from a stolen key or from your own hijacked signing system. Here is what that attacker can and can't do.

Threat model

If this happens, here's what follows.

01One wallet's key is stolenThe attacker spends at most the Instant budget. Anything larger is queued, scored and visible to your team, who can veto it or freeze the vault.
02Your signing system is hijackedSame as a stolen key: the vault, not your signer, enforces the budget and the queue. EIP-7702 delegations to unapproved targets are flagged.
03The attacker tries to add their own walletAdding a wallet needs a majority and a 48-hour timelock, and a pending add makes every payment look more unusual to the Sentinel.
04More than half your wallets are compromisedThen your team has lost control of votes, so keep wallets on different people's devices. The recovery vault and key still can't be changed in under seven days.
05Chainlink CRE stops reportingQueued requests wait the full max delay instead of a scored one. Nothing gets faster; your team can still fast-track or veto.
06A bad report reaches the HubReports are authenticated, bound to one request and single-use. The worst a wrong score does is delay or freeze; it can never send money anywhere.
The admin key

What the palmfi team can do

  • ✓Mark a name verified, or remove the badge
  • ✓Revoke a name (the vault and its funds are untouched)
  • ✓Remove an address from the blocklist
  • ✓Propose a CRE workflow rotation, behind a timelock
The admin key

What it can never do

  • ✕Move, freeze-release or redirect any vault's funds
  • ✕Add or remove a vault's wallets
  • ✕Change a vault's budget, limits or recovery
  • ✕Shorten any delay or override a veto
Audits and reports

Read the work, report what we missed