my private hunting tool (soon to be open-source and @runQuorum played a huge role in helping me trace these
SHIELDS UP
i really hope the attitude of the average crypto users and industry leaders changes after this
i still remember telling my close friends their cryptos are SAFU with binance because the risk of hacks were relatively low and their funds were backed the insurance fund
yet binance got hacked
i really hope this is serves as the final wake-up call for more investments into the neglected security space
hacks such as this damage the already underwhelming reputation of crypto among the masses generally
we need to put and end to them by preventing as much hacks as possible and creating emergency response security teams and alerts to support the current alert systems such as @blockaid_ @ShieldifySec etc
i traced the @binance hack on-chain
~$6.3B left hot wallets used on a daily basis, about $3B is frozen.
binance say they will cover all of it from their insurance fund (SAFU) and company money, withdrawals are currently paused
binance initially pointed at a DevOps engineer’s work PC being the source of the hack, not a stolen private key
it’s now suspected that the hack comes from more than one factor, and that computer may be the entry point.
the full report isn’t out yet
on ETH, i recovered 50 spending permissions signed by binance’s own wallet. only 1 went through.
the other 49 were already out of date.
that same wallet kept sending normal txs during the theft, the attacker didn’t lock them out.
around 00:10 UTC extra spending permission was turned off
a little ETH is still sitting in that hot wallet, the stolen ETH is in the five wallets further down.
the attacker’s wallet that pushed the permissions got funded through a mixer
note that a mixer doesn’t tell you who funded it.
five ETH wallets still hold what left. none of them has sent a tx.
197,460 ETH
0x8532FA14ACAf483729C449b8458aedDb900543D4
137,602 ETH
0xc887aF690E0301CD4edBbCe8F8d31b4EC65B286A
78,415 ETH
0x897C80d1a9729E94460e2Bd93178bD5215158DF3
32,954 ETH
0x54401f80Da97893a15D43566184635B6674BC942
19,740 ETH
0x79e70d139fC05AA75F75611fDF2680E9952B544A
QUORUM retweeted
We pointed Quorum at the 20 largest money-holding contracts on Robinhood Chain.
Together they represented more than $400M.
The result: no exploitable bugs we could confirm.
Most of the code was standard and well built.
That is a result from that run.
It is not a certificate for the next block.
We are working down the list and will publish what we can prove.
Keep up with our updates:
t.me/runQuorumchat
QUORUM retweeted
An upcoming design would let independent operators run checkers and stake to show they are separate.
That design is not in the current tool.
There is no release date year as of now.
Four constraints remain open:
-Checker versions have to be pinned, or results are not reproducible. The required strictness is not settled.
-A weak checker is not a dishonest one. Penalties do not fix low coverage.
-Checkers that are too similar can pass validation and still fail the independence test.
-Below a sufficient number of participants, independence is not meaningful. That threshold is not set.
The current product does not depend on this design. For now, prove it or drop it
runquorum.site/
QUORUM retweeted
Another day to remind you.
The goal is not a scanner that always has something to say.
The goal is a scanner that knows when it has enough reason to speak.
That is Quorum.
Six independent lenses read the contract.
No single lens can publish a finding.
Two have to reach the same conclusion, for different reasons.
Sometimes they do not.
The report stays quiet.
We are trying to make every line on the results mean something.
Still building. Still testing.
runquorum.site
QUORUM retweeted
Last week updates.
We rebuilt Quorum because a flag was too easy to believe.
A bug is not real NOW until the tool writes the attack, runs it on a fork of live state, and watches the money move.
If the money does not move, it stays off the report.
You give it a chain, an address, a block, and a one-line hypothesis.
PROVEN if it pays. UNPROVEN if it does not, with a reason.
It still does not watch the room.
A compromised laptop or a hot wallet drained by a third-party tool stays outside the report.
Bitget was that case. Quorum would not have caught it.
An empty report is not a certificate of safety.
It is the tool refusing to guess.
Prove it or drop it.
runquorum.site
QUORUM retweeted
Your scanner found 20 issues.
How many would you actually trust?
That's something we've been thinking about while building Quorum.
A scanner can flag something suspicious, but that doesn't automatically make it a bug.
So we built Quorum to be stricter.
It has 6 independent lenses. one can't report a bug on its own. two have to agree, for different reasons.
Sometimes, nothing makes it into the report.
We'd rather have a finding earn its place than give you another thing to investigate for no good reason.
still building. still testing. more to share as we go.
runquorum.site
QUORUM retweeted
Another day to inform you that Quorum does not start with a verdict.
It starts with a guess you have to write down.
Give it a chain, an address, a block, and a one-line hypothesis.
It will try to make that hypothesis pay on a fork of live state.
If the money moves, the result is PROVEN.
If it does not, the result is UNPROVEN, with a reason.
A checker having a feeling is no longer the report.
The report is what survived the attempt.
QUORUM retweeted
We rebuilt Quorum because the old reports were too easy to believe.
A checker could flag something, and that was enough to call it a bug. Some of those flags were undeserving.
Now a finding has to survive a harder test. Quorum writes the attack and runs it on a fork of the live chain. If the money does not move, it stays off the report.
An empty result is better than a bug not shown.
Test your own contract:
runquorum.site
the latest on @runQuorum , so many major changes have been made since launch!
it’s now a fully open-source security tool here to serve the community
QUORUM retweeted
The checkers do not talk to each other.
No message. No queue. Nothing passed from one to the next.
Each one writes what it saw into one shared file.
Each one reads what the others wrote.
That file is Sibyl Memory.
The agreement happens there, not inside any checker.
Delete the file and they still see the bugs.
They just never learn that another one saw them, so nothing gets reported.
QUORUM retweeted
Another day, another reminder.
Quorum does not trust the opinion of a single checker, even if it's sure of itself.
It runs 6 checkers with each one reading the contract in its own way.
None of them gets to decide alone.
A bug is reported only when two of them catch it, for two different reasons.
Until then, it is just a checker having a feeling.
Sometimes they do not agree and the report stays empty.
That can look like the tool failed.
It did not. It just refused to guess.
i just submitted @runQuorum to the arbitrum open-house buildathon
runquorum.site
wish us luck guys💚
QUORUM retweeted
Ser why's it called Quorum? Fair question, it wasn't random.
A quorum is the minimum number of people that have to agree before anything counts as a real decision. one person disagreeing means nothing on its own, that's just an opinion.
Same rule runs the scanner. one checker flagging something is just a checker having a feeling. two checkers, catching it for different reasons, actually agreeing, that's when it becomes a finding.
The name wasn't picked after the build. it was the standard we built up to.
it honestly feels nice to have your work appreciated
@runQuorum detected a latent critical bug which was found internally by the @SOCKETProtocol & @BungeeExchange team
here to serve💚
i just submitted another bug i found using @runQuorum, hopefully i get some feedback
it’s a critical but latent bug
updates soon💚
QUORUM retweeted
⚡️ QUORUM 7-DAY CONTEST
7 days of real engagement, visibility & activity.
To qualify:
• Follow @runQuorum
• Join the TG: t.me/runQuorumchat
• Take part in the activities
• Quality > spam
🏆 $50 prize pool
🥇 $25
🥈 $15
🥉 $10
Full details in TG.
QUORUM retweeted
New month. Same rule.
Quorum runs on your machine.
Nothing leaves it.
Clone the repo.
Point it at a contract.
Read the report yourself.
No account.
No key.
A finding means two lenses agreed.
An empty report means they did not.
github.com/Yonkoo11/quorum
QUORUM retweeted
Dev update.
We ran the full pipeline on hacks we had never inspected. Intake to a working exploit proof, end to end.
The scanner found real on-chain victims and reproduced two separate exploits on a fork, draining funds without faked state.
Intake tested clean on three hacks. It captured both proxy and implementation contracts on its own.
The blind run broke the response parser twice on unexpected text. That is fixed. Scans no longer crash on that formatting.
QUORUM retweeted
Dev update:
When target funds sit in connected accounts, the scan used to stop too early.
There is now a search step for those third-party accounts.
The tool checks on-chain state for affected addresses instead of giving up.
We also built two case studies.
Both are backed by working fork proofs from past exploits.
QUORUM retweeted
Dev update:
We pointed the two lenses at source from contracts that were actually hacked.
23 of those attacks matched the shapes the lenses look for.
Only 1 in 9 were caught.
Quorum does not see everything.
An earlier number here was an upper bound, not a hit rate.
The work now is to cut what only helped the demo,
keep what catches real bugs,
and publish the misses with the hits.










