testnetmainnet
LØ

DEX Safety Benchmark

Which decentralized exchanges keep user funds safe — FTB measured against its peers, 0–100 on a fixed 10-element rubric. Below, the Layer-1 each one runs on, scored for decentralization. Open any score breakdown for the elements behind the number. A dated design + track-record posture, not a verdict.

FTB DEXkey status

Every FTB contract and who can still change it, read live from the chain. Keyless — ownership renounced on-chain, immutable forever. Key active — an owner can still call privileged methods. Every address links to the explorer; verify any row yourself.

contractaddresskey
FTB tokenNEP-17 · hard cap 100,000,000—on-chain
LendingFtbLoanBook · P2P term loans · no owner key—Keyless
GovernanceFtbGov · staked-FTB votes—on-chain
Treasuryproposal budgets—on-chain
Bond vaultlisting bonds—on-chain
Dividendsstaker payouts—on-chain
Commandmentson-chain charter—on-chain
Migration claimstateless proof-checker—on-chain
Registryseam board · timelock-governed—on-chain
AMM routerFTB-deployed swap (audited-Flamingo source) · admin via getAdmin() · renounce = burn, not zero—on-chain

How a key becomes keyless

  1. Deploy with a bootstrap key. Each contract is published by a deploy wallet that can only set its launch parameters and wire the registry seams. That key never takes custody — every contract is non-custodial, so funds move only on a transfer you sign in your own wallet.
  2. Lock the parameters. Once the launch values are set and verified on-chain, there is nothing left for an owner to do.
  3. Renounce. The owner calls renounce, which erases the owner slot in the contract's own storage. The transaction is public; the slot reads empty from that block forward, and no address — not the deployer, not any developer, not Free the Bird — can call a privileged method again. The code is frozen as deployed.
  4. Born keyless. The lending contract (FtbLoanBook) ships with no owner method at all: there is nothing to renounce, and it is immutable from its first block.

The table above reads each contract's owner slot live from the chain. Keyless means that slot is empty — confirm it yourself on the explorer. Deployment and development hold no key to the live protocol.

DEX Safetyapp layer · FTB vs peers

Whether any path steals or loses user/pool funds, scored 0–100 on a fixed 10-element rubric (fund safety, oracle integrity, access control, verification depth …). Open score breakdown under any DEX for the elements behind the number. FTB's number is a design projection — unlaunched, $0 at risk — not a track record.

1FTB

Safety84.4

Oracleless, keyless peer-to-peer design — no oracle to manipulate, no owner key, no liquidation price; the willing lender prices the collateral. On oracle integrity and authority it leads every peer on paper — but unlaunched and unproven: not yet live, $0 at risk, 0 days in production. A design projection; the target is to earn the score, not claim it. confidence: low (unlaunched)

score breakdown
Safety — composite 84.4/100 (score × weight ÷ 10)
elementwt/10pts
fund safetyNon-custodial: funds move only on your signed transfer, and the loan book is keyless with proven conservation. The -1: the AMM pool runs under FTB's own pre-launch deploy key — the same key as the rest of the stack, not an outside party — burned on-chain at launch, not a NEO-L1 limit.18916.2
oracle & price integrityFull marks. There is no oracle: the lender prices the collateral and swaps are caller-min-out protected, so the single most-exploited DeFi surface (price-feed manipulation) does not exist here.121012
economic-invariant coverageNo LTV, no liquidation price, no bad-debt cascade; supply<=cap, one-claim and conservation are pinned by firing tests. The -1: proven by tests, not yet by formal proof.12910.8
access-control & authorityLoan book keyless by construction; authority via CheckWitness + GetCallingScriptHash (never tx.sender); mint gated to the migrator. The -1: the rest of the stack still holds pre-launch owner keys pending the renounce (read live above).12910.8
blast-radius containmentA default is a clean collateral->lender swap, no cross-contract cascade. The -2: the swap crosses into the AMM (FTB's own deployed router/factory/pairs), and the stack still holds its pre-launch keys; both close when every key is burned on-chain at launch.1088
verification depthThe honest anchor. Reproducible builds, full public source and the field's incident catalog imported as firing tests are real — but there are no formal proofs, no external audit and 0 days in production. #1 on design only; this is the number time must earn down.1055
feature-surface disciplineOne signing path, a NEO/GAS-only allowlist, no hooks or plugins. The -2: the swap adds the AMM's pair/LP surface — FTB's own deployed contracts, audited-Flamingo source — beyond the minimal core.886.4
signing & front-end integrityWYSIWYS review-and-sign, config-pinned recipients, least-privilege scopes (no blanket approvals), a DOM-sink-free CSP-locked client — tighter than Permit2. The -1: integrity rests on reaching the genuine IPFS build (the own-node control + on-chain re-check cover a tampered mirror).897.2
upgrade & governance safetyThe loan book is immutable (no upgrade path at all); governance is on-chain and keyless. The -2: pre-launch Update/owner keys on the rest of the stack are still live — the biggest closable gap, tracked on the key panel.684.8
residual-risk transparencyResiduals are named in the open: the unlaunched caveat is loud on the card, the key panel reads live, the vuln catalog is public. The -2: full marks want published external review + the renounce transaction ids on record (pending launch).483.2

2Uniswap

Safety83.4

Immutable core, no admin over pool funds, Certora formal proofs, a $15.5M bounty; 2020 (v4 Jan 2025), ~$3.7B TVL, no core-contract loss in 5+ yrs. The earned high bar. confidence: high

score breakdown
Safety — composite 83.4/100 (score × weight ÷ 10)
elementwt/10pts
fund safetyImmutable core, no admin key over pools or user funds; no core-contract theft in 5+ years at multi-billion TVL. The -1 is reserved for the unknowable, not a known hole.18916.2
oracle & price integrityThe -3: the TWAP oracle is manipulable on thin pools — a consumer-integration risk Uniswap documents but cannot prevent.1278.4
economic-invariant coverageBattle-tested constant-product invariants, no bad-debt model. The -1: v4 hooks can inject third-party economic logic.12910.8
access-control & authorityNo admin over funds; governance's only core lever is a capped fee switch (it cannot pause, upgrade or seize). The -1: that switch now directs real fee cash flow.12910.8
blast-radius containmentImmutability caps worst-case damage. The -2: v4 hooks are unvetted third-party code — a 0x study flagged most sampled hooks as malicious or suspect.1088
verification depthBest-in-DeFi: Trail of Bits + Certora formal verification (v4), a $2.35M contest, a $15.5M bounty, 5+ years unexploited. The -1: hooks sit outside the proven core.1099
feature-surface disciplineDisciplined core. The -2: v4 hooks and the periphery widen the surface third parties can misuse.886.4
signing & front-end integrityThe -4: Permit2 blanket approvals are a standing phishing surface that has drained seven figures from users.864.8
upgrade & governance safetyNon-upgradeable core; governance cannot touch pool assets. The -1: Dec-2025 UNIfication points governance at fee cash flow.695.4
residual-risk transparencyMature, well-documented risk disclosure and audits. The -1: hook risk is new and still being characterized.493.6

3PancakeSwap

Safety59.6

The largest BNB-Chain AMM (2020); audited (CertiK, BlockSec, Pashov) with a live bounty, no core-contract theft (the 2021 loss was a DNS / front-end hijack). But contracts are upgradeable behind a timelock + multisig, on Binance-controlled BNB Chain. confidence: medium

score breakdown
Safety — composite 59.6/100 (score × weight ÷ 10)
elementwt/10pts
fund safetyAudited AMM, 5yr with no core theft (the 2021 loss was a DNS hijack, not the contracts). The -4: upgradeable proxies can change fund logic.18610.8
oracle & price integrityThe -4: relies on external/TWAP feeds and third-party integrations; no manipulation-resistant core oracle.1267.2
economic-invariant coverageProven constant-product pools. The -3: a sprawling suite (perps, prediction, lottery) widens the economic surface.1278.4
access-control & authorityThe -5: upgradeable proxies behind a timelock + multisig, on Binance-controlled BNB Chain — real admin authority over the contracts.1256
blast-radius containmentThe -4: a shared proxy-admin and a wide product suite mean one key's reach spans many products.1066
verification depthAudited (CertiK, BlockSec, Pashov) + live bounty + 5yr track record. The -3: no formal proofs, and upgradeability dilutes 'what's audited is what's deployed'.1077
feature-surface disciplineThe -5: a very large feature surface (AMM, perps, prediction, lottery, IFO) — many paths to secure.854
signing & front-end integrityThe -4: standard token-approval UX with the usual blanket-approval phishing exposure.864.8
upgrade & governance safetyThe -5: upgradeable via timelock + multisig — mutable by a small admin set, not renounced.653
residual-risk transparencyThe -4: discloses audits and bounty but not a full live authority/key map.462.4

4Hyperliquid

Safety48.4

Perp DEX on its own L1; custodies margin and the HLP vault. Mar 2025 'validator put' — 5 validators (>50% of stake) delisted JELLYJELLY and force-settled it to cap $10M+ of HLP losses, overriding the market. No core theft, but a set small enough to intervene. confidence: medium

score breakdown
Safety — composite 48.4/100 (score × weight ÷ 10)
elementwt/10pts
fund safetyThe -5: it custodies all margin plus the HLP vault; user funds sit under protocol/validator control.1859
oracle & price integrityThe -5: a validator-set oracle; Mar-2025 showed the set can move a mark to protect the vault.1256
economic-invariant coverageThe -5: a perps/vault model where the house backstop (HLP) can be forced to absorb a position.1256
access-control & authorityThe -6: 5 validators hold >50% stake — a small set can override markets (the JELLY 'validator put').1244.8
blast-radius containmentThe -5: concentrated validator authority means one coordinated action reaches the whole venue.1055
verification depthThe -5: audited and fast-moving, but a novel custodial design with limited independent formal review.1055
feature-surface disciplineThe -5: a full on-chain perps exchange plus vaults — a broad, complex surface.854
signing & front-end integrityThe -4: non-custodial order signing, but the venue settles; the user does not self-custody positions.864.8
upgrade & governance safetyThe -7: a small validator set force-settled a live market by vote — governance can override outcomes.631.8
residual-risk transparencyThe -5: the incident was disclosed, but the custodial/validator-authority model is under-documented for users.452

5THORChain

Safety41.8

Protocol-custodied cross-chain TSS vaults; exploited 2021 (~$16M) & May 2026 (~$10.7M), and a Jan 2025 lending crisis froze ~$200M. Strong containment and incident transparency, but validators have frozen user funds by vote. confidence: medium

score breakdown
Safety — composite 41.8/100 (score × weight ÷ 10)
elementwt/10pts
fund safetyThe -7: the protocol custodies all pool funds in TSS vaults — drained twice (2021 ~$16M, May-2026 ~$10.7M).1835.4
oracle & price integrityThe -5: cross-chain price/accounting via node consensus; forged deposit events caused the 2021 loss.1256
economic-invariant coverageThe -7: the Jan-2025 THORFi lending crisis froze ~$200M and forced debt tokenisation — an economic near-death.1233.6
access-control & authorityThe -5: node operators tune live 'Mimir' params; a privileged Admin-Mimir override reportedly still exists.1256
blast-radius containmentThe -6: a single vault-key compromise drains that vault; chain-halts are network-wide.1044
verification depthThe -5: audited + a $500k bounty, but the exploited TSS/soundness paths were not caught pre-incident.1055
feature-surface disciplineThe -7: swaps + savers + lending + synths across many chains — a very wide custodial surface.832.4
signing & front-end integrityThe -4: non-custodial signing across independent front-ends — a genuine strength.864.8
upgrade & governance safetyThe -7: validators froze user funds by emergency vote; live params are mutable by design.631.8
residual-risk transparencyThe -3: strong, prompt incident post-mortems and transparency — its best dimension.472.8

6Rhea Finance

Safety35.6

NEAR's largest DeFi hub (Ref + Burrow merger; >95% of NEAR TVL at peak). Apr 2026: a fake-token oracle attack distorted price feeds and drained ~$7.6M (CertiK); withdrawals paused. The exact failure FTB's oracleless design removes. confidence: medium

score breakdown
Safety — composite 35.6/100 (score × weight ÷ 10)
elementwt/10pts
fund safetyThe -7: an Apr-2026 fake-token oracle attack drained ~$7.6M; withdrawals were paused.1835.4
oracle & price integrityThe -8: the direct cause of death — price feeds distorted by a fake token. The exact class FTB's oracleless design removes.1222.4
economic-invariant coverageThe -7: a lending/AMM model exposed to feed manipulation and share distortion.1233.6
access-control & authorityThe -6: pausable with admin controls (withdrawals were halted post-hack) — real operator authority.1244.8
blast-radius containmentThe -6: a merged Ref + Burrow stack means one bad feed reaches lending and swaps together.1044
verification depthThe -6: audited (CertiK) yet the oracle path failed in production — the audit did not catch it.1044
feature-surface disciplineThe -6: NEAR's largest DeFi (>95% of TVL) spanning AMM + lending — a wide surface and a single point of systemic risk.843.2
signing & front-end integrityThe -5: standard web-dapp signing; no notable hardening.854
upgrade & governance safetyThe -7: admin-pausable and upgradeable; authority is concentrated.631.8
residual-risk transparencyThe -4: disclosed the incident and paused promptly; partial transparency.462.4

7Flamingo

Safety27.8

Upgradeable contracts whose keys the team says sit with Neo's foundations; two 2026 exploits (one minted ~2.19 trillion FLM); TVL ~$0.6M, near-dead. confidence: medium

score breakdown
Safety — composite 27.8/100 (score × weight ÷ 10)
elementwt/10pts
fund safetyThe -8: it custodies pooled funds; an Aug-2026 bug minted ~2.19 trillion FLM and drained the pool — a total loss.1823.6
oracle & price integrityThe -7: price/economic paths proved manipulable; the bounty even excludes oracle/economic bugs.1233.6
economic-invariant coverageThe -8: a staking<->lend interaction minted ~3,500x supply — a total economic-invariant failure.1222.4
access-control & authorityThe -8: upgradeable contracts whose keys, per Flamingo's own letter, sit with Neo's foundations — nothing renounced.1222.4
blast-radius containmentThe -7: shared upgrade authority plus heavy bNEO/Poly-bridge dependencies — failures cascade.1033
verification depthThe -6: historically audited (NCC/CertiK/PeckShield) but the exploited paths were not caught; the bounty carves out the risky classes.1044
feature-surface disciplineThe -7: full-stack (Swap/Lend/Perp/legacy Flamincome) — the largest attack surface here.832.4
signing & front-end integrityThe -6: standard dapp signing; no notable client hardening.843.2
upgrade & governance safetyThe -8: live external admin keys, nothing renounced, now tangled in a Neo co-founder governance dispute.621.2
residual-risk transparencyThe -5: public source and audit history, but the bounty exclusions understate the real residual risk.452
Chainsthe L1 each DEX runs on

A DEX's safety is capped by its chain. These are the Layer-1s the DEXes above run on, scored for decentralization (and quantum resistance, for context). The full quantum-resistance benchmark across the wider field is archived for the Libration Zero launch.

NEONEO · FTB base

Decentr.39.6
Quantum9.4
Privacy—

dBFT L1 — FTB's current base layer. Fast single-block finality.

WeaknessesClassical ECDSA secp256r1, no PQ plan; 7-node consensus; the committee can block accounts on-chain. An app on NEO inherits all of this.

score breakdown
Quantum resistance — composite 9.4/100 (score × weight ÷ 10)
elementwt/10pts
signature-scheme class1800
default vs dormant1400
maturity & standardization1200
independent audit1000
cryptanalytic params1000
public-key exposure1011
hash/Grover margin663.6
harvest-now-decrypt-later854
migration / Q-day plan810.8
hybrid / fallback400
Decentralization — composite 39.6/100 (score × weight ÷ 10)
elementwt/10pts
consensus set & entry1646.4
vote/stake concentration (Nakamoto)1434.2
governance mutability1434.2
censorship & halt1222.4
upgrade path1055
client diversity875.6
node count & spread854
treasury/foundation control832.4
mempool/MEV653
finality trust462.4

Privacy — rubric defined, scoring in progress.

EthereumETH

Decentr.61.8
Quantum14.2
Privacy—

PoS smart-contract L1; ~1M validators and the strongest client diversity among the majors.

WeaknessesClassical ECDSA (funded PQ plan ~2029); Lido + CEX staking concentration (entity Nakamoto ~3); large MEV economy.

score breakdown
Quantum resistance — composite 14.2/100 (score × weight ÷ 10)
elementwt/10pts
signature-scheme class1800
default vs dormant1400
maturity & standardization1211.2
independent audit1000
cryptanalytic params1000
public-key exposure1011
hash/Grover margin663.6
harvest-now-decrypt-later854
migration / Q-day plan854
hybrid / fallback410.4
Decentralization — composite 61.8/100 (score × weight ÷ 10)
elementwt/10pts
consensus set & entry16711.2
vote/stake concentration (Nakamoto)1445.6
governance mutability1479.8
censorship & halt1267.2
upgrade path1066
client diversity875.6
node count & spread886.4
treasury/foundation control864.8
mempool/MEV642.4
finality trust472.8

Privacy — rubric defined, scoring in progress.

BNB ChainBNB

Decentr.28.6
Quantum11.8
Privacy—

EVM proof-of-stake chain; ~45 validators, high throughput — the home of PancakeSwap.

WeaknessesClassical ECDSA, no PQ plan; Binance-dominated validator set (entity Nakamoto ~7); validators halted the chain after the Oct 2022 Token Hub bridge hack (~$570M).

score breakdown
Quantum resistance — composite 11.8/100 (score × weight ÷ 10)
elementwt/10pts
signature-scheme class1800
default vs dormant1400
maturity & standardization1211.2
independent audit1000
cryptanalytic params1000
public-key exposure1011
hash/Grover margin663.6
harvest-now-decrypt-later854
migration / Q-day plan821.6
hybrid / fallback410.4
Decentralization — composite 28.6/100 (score × weight ÷ 10)
elementwt/10pts
consensus set & entry1634.8
vote/stake concentration (Nakamoto)1434.2
governance mutability1434.2
censorship & halt1222.4
upgrade path1033
client diversity821.6
node count & spread832.4
treasury/foundation control821.6
mempool/MEV642.4
finality trust452

Privacy — rubric defined, scoring in progress.

HyperliquidHYPE

Decentr.23.6
Quantum11
Privacy—

Purpose-built orderbook / perp L1 (HyperBFT); sub-second finality — the home of the Hyperliquid DEX.

WeaknessesClassical crypto, no PQ plan; ~16–21 validators with 5 holding >50% of stake (Hyper Foundation); validators have overridden live markets (the JELLY 'validator put').

score breakdown
Quantum resistance — composite 11/100 (score × weight ÷ 10)
elementwt/10pts
signature-scheme class1800
default vs dormant1400
maturity & standardization1211.2
independent audit1000
cryptanalytic params1000
public-key exposure1011
hash/Grover margin663.6
harvest-now-decrypt-later854
migration / Q-day plan810.8
hybrid / fallback410.4
Decentralization — composite 23.6/100 (score × weight ÷ 10)
elementwt/10pts
consensus set & entry1623.2
vote/stake concentration (Nakamoto)1422.8
governance mutability1422.8
censorship & halt1222.4
upgrade path1022
client diversity821.6
node count & spread832.4
treasury/foundation control821.6
mempool/MEV642.4
finality trust462.4

Privacy — rubric defined, scoring in progress.

THORChainRUNE

Decentr.41
Quantum9.4
Privacy—

Cross-chain swap L1 (Tendermint BFT, bonded nodes). Also scored as a DEX below.

WeaknessesClassical secp256k1 TSS; ~100 bonded nodes with an Admin-Mimir override; froze user funds in 2025; single client.

score breakdown
Quantum resistance — composite 9.4/100 (score × weight ÷ 10)
elementwt/10pts
signature-scheme class1800
default vs dormant1400
maturity & standardization1200
independent audit1000
cryptanalytic params1000
public-key exposure1011
hash/Grover margin663.6
harvest-now-decrypt-later854
migration / Q-day plan810.8
hybrid / fallback400
Decentralization — composite 41/100 (score × weight ÷ 10)
elementwt/10pts
consensus set & entry1658
vote/stake concentration (Nakamoto)1457
governance mutability1434.2
censorship & halt1233.6
upgrade path1044
client diversity821.6
node count & spread854
treasury/foundation control843.2
mempool/MEV653
finality trust462.4

Privacy — rubric defined, scoring in progress.

NEARNEAR

Decentr.48.8
Quantum11.8
Privacy—

Sharded PoS L1 (Nightshade); ~200+ validators — the home of Rhea Finance and NEAR Intents.

Weaknessesed25519 (classical), no PQ plan; a single dominant client; the NEAR Foundation retains significant stake and influence.

score breakdown
Quantum resistance — composite 11.8/100 (score × weight ÷ 10)
elementwt/10pts
signature-scheme class1800
default vs dormant1400
maturity & standardization1211.2
independent audit1000
cryptanalytic params1000
public-key exposure1011
hash/Grover margin663.6
harvest-now-decrypt-later854
migration / Q-day plan821.6
hybrid / fallback410.4
Decentralization — composite 48.8/100 (score × weight ÷ 10)
elementwt/10pts
consensus set & entry1658
vote/stake concentration (Nakamoto)1457
governance mutability1457
censorship & halt1256
upgrade path1055
client diversity832.4
node count & spread864.8
treasury/foundation control843.2
mempool/MEV653
finality trust462.4

Privacy — rubric defined, scoring in progress.

verify verify everything - trust nothing v0.11
contracts — check each against the repo
token
-
migrator
-
registry
-
ballots
-
integrity
merkle root
-
max supply
-
mint sealed
-
snapshot
block 12,782,704 · reproducible with snapshot_state.py against any archive node

check every hash against the repo before you sign. after close + renounceOwnership the token has no admin — readable on-chain, forever.