Skip to content

SECURITY.md

Latest commit

 

History

History
97 lines (70 loc) · 4.24 KB

SECURITY.md

File metadata and controls

97 lines (70 loc) · 4.24 KB

OverSync v2 — Security

This document is the OverSync v2 threat model, audit preparation checklist, and bug bounty plan. It deliberately calls out the things we have NOT yet done, so any reader can verify each claim.

Status

Asset Audit status Notes
soroban/contracts/htlc Unaudited. Has 10 unit tests. Slated for independent audit pre-mainnet (Tranche 2 deliverable).
soroban/contracts/resolver-registry Unaudited. Same plan as above.
contracts/v2/HTLCEscrow.sol Unaudited. 15 Hardhat tests + slither lint in CI. Audit prep includes Foundry fuzz + invariant suite (Tranche 1 deliverable).
contracts/v2/ResolverRegistry.sol Unaudited. 6 Hardhat tests. Owner role intended for multisig before mainnet.
Coordinator + SDK + frontend Out of scope for security audit — they cannot move user funds (see TRUST_MODEL.md). Static analysis (eslint + tsc strict) only.

Threat model

We use a STRIDE-style breakdown.

Spoofing

Threat Mitigation
Attacker impersonates the coordinator API to phish users Coordinator URL is hardcoded in the frontend bundle; users are educated to verify the published oversync.app domain.
Attacker impersonates a resolver Resolvers are identified by their on-chain stake in the registry; a non-staked address simply cannot be matched against isActive.

Tampering

Threat Mitigation
Coordinator DB is corrupted DB is a cache; it can be rebuilt from on-chain events. No user funds depend on it.
Frontend bundle is replaced via DNS hijack Standard mitigation: CSP, SRI on critical assets, build provenance via GitHub releases.

Repudiation

All state-changing actions on the bridge happen via signed on-chain transactions visible on either Etherscan or Stellar Expert.

Information disclosure

The bridge does not handle PII. Order metadata (addresses, amounts, hashlocks) is public on-chain by design.

Denial of service

Threat Mitigation
Coordinator DDoS Rate-limit + Cloudflare in front of the public deployment. Even with the coordinator offline, users can still refund directly from contracts.
Resolver collusion to ignore an order Anyone can run a resolver; users can also self-resolve by participating as their own resolver.
Public RPC rate-limits during high traffic Multiple Alchemy / QuickNode endpoints in a round-robin pool.

Elevation of privilege

Threat Mitigation
Admin can drain HTLC contracts The HTLC contract has no admin role with fund-moving authority (no emergencyWithdraw, no pause, no transferOwnership). Verified in the test non-custodial guarantees > contract has no admin escape hatch.
Registry admin can drain resolver stakes outside of slashing slash is the only privileged action and it routes funds to slashBeneficiary, not to the admin EOA.

Audit preparation checklist

Pre-audit (Tranche 1):

  • Single canonical EVM HTLC contract (HTLCEscrow.sol)
  • Single canonical Soroban HTLC contract (oversync-htlc)
  • No admin escape hatches in HTLC contracts
  • Reentrancy guards on every state-changing function (OZ ReentrancyGuard)
  • SafeERC20 on every token transfer
  • OpenZeppelin v5 used (Ownable2Step for the registry)
  • 10 Soroban unit tests + 21 Hardhat unit tests in CI
  • Foundry fuzz + invariant tests (planned)
  • Slither must-not-fail CI gate (currently advisory)
  • Differential testing: same hashlock works on both chains

Audit (Tranche 2):

  • Engage two independent auditors for the HTLC contracts
  • Public audit reports + remediation diff
  • Bug bounty announced

Bug bounty

We will open a public Immunefi-style bounty once both HTLC contracts are audited. Until then, please email security findings to security@oversync.app (PGP key to be published). We commit to:

  • Acknowledging within 48h.
  • Crediting reporters on the SECURITY.md release notes (unless they prefer anonymity).
  • Not pursuing legal action against good-faith security research that follows responsible disclosure.

Out of scope for v2

  • Optimistic rollup support
  • Native Bitcoin support
  • Off-chain MEV mitigation beyond hashlock + timelock semantics