Skip to content
Wall 01 See

Live Detection

Upstream of finality, where the number can still change. It reads every chain in the registry, and the coverage page says what each has earned rather than implying parity. It looks at what is being attempted rather than only what succeeded, and simulates outcomes rather than pattern-matching signatures. When it fires it says what it saw, how confident it is, and what it is worth.

Wall01 of 05Watches5 surfacesBuilt for4 groupsLosses answered5Refuses a transactionYes, once you arm it

How it works.

One core serves four lanes, and they share what they learn: an address that drained a wallet user is already known to the protocol lane. Rules and simulation are the spine. Rather than matching a transaction against known-bad shapes, the system runs it against a copy of current chain state and reads what it would really do. Machine learning ranks what to look at first; it never blocks or trips a breaker alone. Findings land on a seven-rung ladder, from observe to halt, and you set the ceiling above which nothing can fire.

  • Pre-inclusion visibility
  • Outcome simulation
  • Seven-rung response ladder
  • The Circuit Breaker, owned by you

What it actually looks at.

Five surfaces, named. Anything outside this list is not covered by this wall; the perimeter page says which one covers it.

What a transaction is attempting, run against current chain state to see what it would really do
The pending pool, and the moment before a wallet user approves anything
Every EVM chain in the registry, read into one model and one threat graph, so an attacker who moves between them stays one incident rather than becoming several
A ceiling you set, above which no automatic response can ever fire
An expiry on every signed alert, so a stale finding cannot be replayed against your contract next year

Who it is for.

Protocol lane

Minutes to start. No code, no contract, and no cooperation needed from anyone else.

Pick the preset for what you are, and findings arrive on the channel your team already reads. Before you commit, the preset states in writing where its boundary sits: an issuer whose reserves are attested off chain is told that plainly rather than handed a green tick. A bridge belongs here too, because it configures contracts, escrow and message correspondence like any other protocol. The Circuit Breaker is available on this lane.

  • DeFi protocols
  • DAOs and protocol treasuries
  • Corporate treasuries
  • Custodians, funds and asset managers
  • Stablecoin issuers
  • Liquid staking and restaking protocols
  • RWA issuers
  • Perpetual venues
  • Cross-chain bridges

Exchange lane

A server-side call on the deposits, withdrawals and counterparties you already process.

You get a decision with the sources that produced it. An address we have never seen returns insufficient-data, never a pass, because an empty answer and a clean answer are not the same thing. Every result names which sources were live, so your risk team can tell whether a clear rests on one stale list or on three that agree. The Circuit Breaker covers the contracts you operate.

  • Centralised exchanges
  • Payment and settlement rails
  • OTC and market-making desks

Wallet lane

One call before you show a confirmation screen. No contract, nothing on chain.

You get back allow, warn or block, with a reason you can put on screen: this contract was deployed eleven minutes ago, this approval has no limit, this signature drains your position. Past a hard deadline the SDK degrades to warn. Never to allow, because a verdict nobody could compute is not evidence of safety; never to block, because that would brick every signature during an outage. No Circuit Breaker here: a signing screen has no contract to guard, and we hold no keys.

  • Wallet providers
  • Browser extensions
  • Mobile wallets
  • Smart-account and account-abstraction providers
  • dApp front ends
  • Custody and treasury interfaces
  • Anything with a signing screen

Rollup lane

Screening before inclusion, so a bad transaction never enters a block you produced.

You run the sequencer, so you can decline to include a transaction rather than react afterwards. Screening runs on the ordering you control. The addresses your chain depends on, the inbox, the messenger, the system predeploys, sit on a never-exclude list, because excluding one on a bad label would break your own chain. Everything the protocol lane does applies here.

  • L2 sequencers and rollup operators
  • Appchain and L3 operators

This is the wall that can refuse.

Every other wall reports. This one also carries the contract that declines the transaction, and it belongs to you. It has a page of its own, because how it is armed matters more than that it exists.

The brake is a contract you own.

Fuga Labs signs a statement and never touches it. Your policy contract reads that statement and decides, which is why a total compromise of us buys one alert and nothing else.

How the Circuit Breaker is armed
Two mechanisms, one fast
A revert inside your own transaction, needing nothing from us, and a policy that tightens after a finding. Only the first is fast enough for a one-block exploit.
Four rungs to armed
Disarmed, shadow, canary, armed. Neither publishing rung acts until a shadow period has earned it, and the kill switch returns it in one transaction you hold.

Five losses this wall answers.

Public, dated and sourced. Open a row for what happened and for what would have caught or refused it.

$$285MM Drift Protocol April 2026 · Protocol lane withdrawn in twelve minutes, across thirty-one transactions
What happened

Attackers spent months cultivating the team, then used Solana durable nonces to get Security Council members to pre-sign transactions handing over administrative control. They whitelisted a worthless token as collateral at an invented price, deposited 500 million of it, and withdrew real USDC, SOL and ETH. The withdrawal ran twelve minutes.

What answers it

Thirty-one transactions against one protocol in twelve minutes is a rate, and a rate is what a velocity rule is for. Detection fires on the first few; the armed Circuit Breaker caps outflow per window at a ceiling the treasury set, so the rest revert at your own entrypoint.

It takes more than one wall. Ops Monitor sees the administrative handover as it happens, and an Asset Ratings floor read at collateral onboarding refuses a token with no verifiable structure. Any one narrows this; all three stop it.

TRM Labs on the Drift Protocol heist
$$128MM Balancer V2 November 2025 · Protocol lane drained across six chains in under thirty minutes, then copied onto forks
What happened

A rounding inconsistency in the pool invariant maths let an attacker suppress pool-token prices and extract value through repeated arbitrage. A single swap loses a negligible amount; sixty-five batched into one transaction compound into a drain. It ran across six networks in under half an hour, and copycats hit unpatched forks.

What answers it

One transaction of sixty-five internal operations that leaves the pool worse off is a value-conservation failure, checked inside the transaction rather than alerted after it. The Circuit Breaker reverts it at the venue's own entrypoint. The shared threat graph means a pattern seen on one deployment is already known to every fork.

Check Point Research on the Balancer drain
$$223MM Cetus Protocol May 2025 · Protocol lane taken in under fifteen minutes from a one-token deposit
What happened

An overflow check in a shared maths library failed to reject values above its limit. The attacker flash-loaned, opened a position in a very narrow price range, and added one unit of a token. The overflow reset the calculation to a far smaller number, crediting liquidity worth millions against a deposit of one.

What answers it

The check that matters is arithmetic and local: liquidity minted must be consistent with tokens actually supplied, in the same transaction. Stated as an invariant on the venue's entrypoint, the Circuit Breaker reverts the mint. No off-chain pipeline is fast enough here, and we do not claim one is.

Halborn on the Cetus hack
5HOURSHOURS Ledger Connect Kit December 2023 · Wallet lane with a wallet drainer live inside hundreds of legitimate dApps
What happened

A former employee's package account was phished, and the attacker published malicious versions of a library loaded by hundreds of dApp front ends. The injected code asked users to sign approvals and transfers routed to the attacker. Every dApp was legitimate, every domain correct, every hardware wallet working as designed.

What answers it

There is exactly one place to catch this: between confirm and signature, by reading what the transaction actually does. A pre-signature verdict returns block on an unlimited approval to an address first seen minutes ago, with a reason the wallet renders in plain language.

Ledger's own security incident report
$$197MM Euler Finance March 2023 · Protocol lane beginning to end, inside a single transaction
What happened

An upgrade added a function that reduced the caller's own collateral without running the solvency check every other fund-moving function ran. The attacker flash-loaned, used the donation to push their position into insolvency, then self-liquidated it at a discount and collected the difference.

What answers it

No alerting pipeline can interrupt one transaction: there is no gap between the first step and the last to send a message into. The only thing that stops this is a check inside the same transaction, in a contract the venue itself calls. That is the Circuit Breaker, and it works with Fuga Labs offline.

Veridise on the Euler Finance insolvency vulnerability

What it builds on.

Shared substrate, not features of this wall. Every layer below is already paid for by the first wall in your scope.

Live Detection This wall
one detection core, four lanes: protocol and bridge, exchange, wallet, and rollup sequencer
a shared threat graph every lane writes and every lane reads
transaction simulation on forked state, exercised against a real node
EVM and Solana ingestion canonicalised into one model; the EVM lanes are what a chain earns its way into, and Solana's own indexer is not there yet
custody-free by construction: Fuga Labs observes and attests, every breaker executes through a module you own

How it is operated.

How this wall is run Stated before anything is provisioned

Deterministic first: rules and simulation are the spine, and machine learning informs prioritisation without ever blocking a transaction or tripping a breaker on its own. Every finding lands on a seven-rung ladder set by confidence and stake, and you set the ceiling. Signed alerts carry a mandatory expiry. The Circuit Breaker is opt-in, ships disarmed, and rehearses report-only until the bar you set is met.

What turns on when it is in scope.

Individually, or bundled with the walls beside it.

In scope

All four lanes

Protocol, treasury and bridge monitoring with no code required; deposit and withdrawal screening for an exchange; a pre-signature verdict for anyone with a signing screen; pre-inclusion screening if you run a sequencer

In scope

Refusal you own

The policy contract is deployed by you, configured by you and revocable by you in one transaction. It works with Fuga Labs completely offline

It is the sensor layer. Exposure Map turns its findings into a blast radius, Asset Ratings sets which assets deserve tighter thresholds, Ops Monitor covers what it structurally cannot see, and its bounded-authority contracts are the ones that bound an AI agent.

Fuga Labs is read-and-attest, never custody-and-execute.
The cardinal ruleEnforced in code across every wall, Live Detection included

Put Live Detection on your perimeter.

A scoping call maps this wall against what you run and says where it reaches and where it stops. Nothing is provisioned until that is agreed in writing.