4 ms·

Is there a detailed technical description of this attack somewhere? The "official" page is very light on details (they got 5 validator private keys, no biggie,

Is there a detailed technical description of this attack somewhere? The "official" page is very light on details (they got 5 validator private keys, no biggie, but something..something...rpc whitelist and oh wow now they got all the money). Since 4 of the 5 compromised keys were under the control of one organization, it seems reasonable to wonder about an inside job.

caoilte 5y ago
I can't recommend "web3 is going great" highly enough for creating digestible explanations. There's good commentary on Axie Infinity (the Sky Mavis game which got hacked) and links to more detailed explanations here. https://web3isgoinggreat.com/?id=2022-03-29-0 https://web3isgoinggreat.com/?id=2022-03-29-0 (particularly by Molly White the owner of the web3isgoinggreat https://blog.mollywhite.net/axie-hack/ https://blog.mollywhite.net/axie-hack/ )
- dboreham 5y ago
  Hmm...thanks, but reading those I don't see any more details. There is some good background information on side chains and bridges but I'm already familiar with the security risks in the typical trusted bridge design. For clarity: I'm interested in answers to questions such as : how exactly did the attacker gain control of 5 validator private keys? The validators were running a vulnerable version of the Linux kernel, and the private keys were sitting in the filesystem, for example, could be an answer. And: what exactly is the mumble about RPC about? Did they build some sort of master key into the bridge contract, then left that key sitting on a machine with no root password?
  - phphphphp 5y ago
    Valid questions but they’re answers we will likely never know, because either the hack was an inside job and disclosure is bad for them or because they’re too incompetent to ever identify the real cause. Disclosure beyond “we were hacked” is not common in web3.
    - dboreham 5y ago
      They're pretty important questions imho because if you can just compromise 2/3 the validators in a BFT network "like that" then the whole idea of BFT-based chains is broken.
      - renewiltord 5y ago
        It is interesting that in a 9-validator design, 4 of them were really one with 4 votes. With a 5 of 9 min, that makes this a 2 of 5 design really.
  - caoilte 5y ago
    haha. fair. That's more detail than I want to know.
    - t0mmyb0y 5y ago
levischoen 5y ago
https://rekt.news/ronin-rekt/ https://rekt.news/ronin-rekt/
FatalLogic 5y ago
>4 of the 5 compromised keys were under the control of one organization, And it appears the final key was also under the control of the same organization temporarily. I don't have more details because I don't think they've released any, but fundamentally, that appears to be the critical vulnerability which made this hack almost inevitable. They began with a system that was extremely carefully designed to be decentralized (multiple signers) and then they centralized it. Subsequently, the hacker only had to attack one organization, instead of five Incompetence? Inside job? Or why not both?
carlosdp 5y ago
The quick version: - The bridge was controlled by a 5 of 9 validator set (ie. 5 of the 9 nodes needed to agree to withdrawals and deposits between Ethereum and Ronin) - 4 of these nodes were run by Sky Mavis themselves (a bad idea), those were compromised because their internal infra was compromised in some traditional social engineering attack (ie. not blockchain related, just a normal cloud infra compromise) - 1 of the nodes was run by the community-operated Axie DAO. However, that node had whitelisted Sky Mavis' cloud infrastructure a while back to allow them to run transactions through Axie DAO's node in order to use Axie DAO treasury funds to pay gas on user's behalf (to help with onboarding into Ronin). This was discontinued, but they forgot to remove the whitelist (another critical mistake), so the hackers were able to use that Axie DAO node for the final 5th verification needed to steal bridge funds (because they had compromised the whitelisted Sky Mavis cloud infra). TLDR: A traditional cloud infra compromise via unspecified social engineering attack, compounded by an insufficiently decentralized validator set allowing the bridge funds to be "withdrawn" to hacker's wallets by forging fake signatures. This is why decentralization is important in blockchain infra, and they learned the hardest way possible. Hope that helps!
- dboreham 5y ago
  To honest I don't see how this can be anything other than an inside job. No reasonable person would expect someone to have setup the arrangement where a key controlling a kind of faucet would also be a validator key. That's insane and obviously a very bad idea. So whoever pulled off the attack must have known that a very unusual and very stupid arrangement had been created. Ergo they must have been an insider.
  - carlosdp 5y ago
    I don't think it was an inside job, it's actually pretty hard to wash $600m+ of blockchain funds already (it's highly traceable). Doing it as an insider would be truly insane. The whitelisting wasn't a secret. Axie DAO is an open community, so it's easy to have known the whitelist existed and was left in. It was a very bad idea, but what can I say, people get cocky and make mistakes. The only difference in crypto is that it has a built-in bug bounty of losing stored funds like this, so you have to take security incredibly seriously, and they didn't.
Tao331 5y ago

caoilte 5y ago

I can't recommend "web3 is going great" highly enough for creating digestible explanations. There's good commentary on Axie Infinity (the Sky Mavis game which got hacked) and links to more detailed explanations here. https://web3isgoinggreat.com/?id=2022-03-29-0 https://web3isgoinggreat.com/?id=2022-03-29-0 (particularly by Molly White the owner of the web3isgoinggreat https://blog.mollywhite.net/axie-hack/ https://blog.mollywhite.net/axie-hack/ )

dboreham 5y ago
Hmm...thanks, but reading those I don't see any more details. There is some good background information on side chains and bridges but I'm already familiar with the security risks in the typical trusted bridge design. For clarity: I'm interested in answers to questions such as : how exactly did the attacker gain control of 5 validator private keys? The validators were running a vulnerable version of the Linux kernel, and the private keys were sitting in the filesystem, for example, could be an answer. And: what exactly is the mumble about RPC about? Did they build some sort of master key into the bridge contract, then left that key sitting on a machine with no root password?
- phphphphp 5y ago
  Valid questions but they’re answers we will likely never know, because either the hack was an inside job and disclosure is bad for them or because they’re too incompetent to ever identify the real cause. Disclosure beyond “we were hacked” is not common in web3.
  - dboreham 5y ago
    They're pretty important questions imho because if you can just compromise 2/3 the validators in a BFT network "like that" then the whole idea of BFT-based chains is broken.
    - renewiltord 5y ago
      It is interesting that in a 9-validator design, 4 of them were really one with 4 votes. With a 5 of 9 min, that makes this a 2 of 5 design really.
- caoilte 5y ago
  haha. fair. That's more detail than I want to know.
  - t0mmyb0y 5y ago

levischoen 5y ago

https://rekt.news/ronin-rekt/ https://rekt.news/ronin-rekt/

FatalLogic 5y ago

>4 of the 5 compromised keys were under the control of one organization, And it appears the final key was also under the control of the same organization temporarily. I don't have more details because I don't think they've released any, but fundamentally, that appears to be the critical vulnerability which made this hack almost inevitable. They began with a system that was extremely carefully designed to be decentralized (multiple signers) and then they centralized it. Subsequently, the hacker only had to attack one organization, instead of five Incompetence? Inside job? Or why not both?

carlosdp 5y ago

The quick version: - The bridge was controlled by a 5 of 9 validator set (ie. 5 of the 9 nodes needed to agree to withdrawals and deposits between Ethereum and Ronin) - 4 of these nodes were run by Sky Mavis themselves (a bad idea), those were compromised because their internal infra was compromised in some traditional social engineering attack (ie. not blockchain related, just a normal cloud infra compromise) - 1 of the nodes was run by the community-operated Axie DAO. However, that node had whitelisted Sky Mavis' cloud infrastructure a while back to allow them to run transactions through Axie DAO's node in order to use Axie DAO treasury funds to pay gas on user's behalf (to help with onboarding into Ronin). This was discontinued, but they forgot to remove the whitelist (another critical mistake), so the hackers were able to use that Axie DAO node for the final 5th verification needed to steal bridge funds (because they had compromised the whitelisted Sky Mavis cloud infra). TLDR: A traditional cloud infra compromise via unspecified social engineering attack, compounded by an insufficiently decentralized validator set allowing the bridge funds to be "withdrawn" to hacker's wallets by forging fake signatures. This is why decentralization is important in blockchain infra, and they learned the hardest way possible. Hope that helps!

dboreham 5y ago
To honest I don't see how this can be anything other than an inside job. No reasonable person would expect someone to have setup the arrangement where a key controlling a kind of faucet would also be a validator key. That's insane and obviously a very bad idea. So whoever pulled off the attack must have known that a very unusual and very stupid arrangement had been created. Ergo they must have been an insider.
- carlosdp 5y ago
  I don't think it was an inside job, it's actually pretty hard to wash $600m+ of blockchain funds already (it's highly traceable). Doing it as an insider would be truly insane. The whitelisting wasn't a secret. Axie DAO is an open community, so it's easy to have known the whitelist existed and was left in. It was a very bad idea, but what can I say, people get cocky and make mistakes. The only difference in crypto is that it has a built-in bug bounty of losing stored funds like this, so you have to take security incredibly seriously, and they didn't.

Tao331 5y ago