Where this logic fails:
Controlling 1/3 of the network for 1 blob does not imply
the cost to corrupt EigenDA ~= 1/3 of the ETH stake
Because there is no DAS and intersubjective forking/slashing the attack could work with 0 consequences.
With slashing you need to buy 1/3 of the stake to control 1/3 of the validators, without slashing you don't need to buy anything and just bribe them.
And that bribe does not have to be 1/3 of the stake.
Even the coordination of the attack can be made permissionless as well. Strawman where disperser is malicious:
- Validators run TEE binary
- Disperser makes on-chain request for an attack
- Disperser encrypts payload with TEE Session key
- Validators decrypt payload inside TEE sign it and then delete the data (no availability)
- Validators can claim an onchain payment through verification of a remote attestation
- put it on a privacy layer to also have 0 validator reputation cost
There is probably an easier way to do this but this strawman "guarantees" that the Validators don't have the data, Ethereum think it's available and the user funds on the rollup are at risk for minimal cost as you only need to withhold 1 blob.
Just to be clear I don't advocate doing this and this is just a hypothetical but it is foolish to think that the cost of the attack is more than 1/3 of the Ethereum stake. Having 1/3 of the stake just guarantees it but it can happen much earlier.
TLDR - increases the cost to corrupt/sybil resistance, EigenDA's ETH quorum is one of the default two (ETH & EIGEN) which adds to the cost to corrupt that with the other quorums the blob was sent to (4.3M ETH / 1/3 = 1.43M ETH = ~$2.66B) EIGEN quorum.
Disincentives be supplemented with Proofs of Custody slashing.
~~~~
Sure, in general these blockchains/distributed systems have the concept of a safety threshold.
The safety threshold essentially refers to minimal threshold of participants in the network to be honest. For ETH or other blockchains secured by common consensus algorithm's (e.g. Cosmos's CometBFT)
The network remains secure as long as at least 2/3 of validators (weighted by stake) are honest
The inverse of this means controlling 1/3 of the network can lead to malicious or incorrect action.
So..
1/3 of the stake ~= the cost to corrupt the system or DA network in this case.
EigenDA structure roughly looks like this
- All validators have stake to participate in EigenDA and opt into quorums, min is 32 ETH in the ETH quorum
- Data is sharded across the network of validators
- Validators validate, store data and send signatures back
- Once a sufficient threshold (above this safety threshold across ALL quorums) you have an attestation the data is stored and will be stored
So for EigenDA this means the safety threshold is ETH Quorum
- 4.3M ETH * 1/3 = 1.43M ETH ~$2.66B (@ $2,000 ETH)
- EIGEN quorum value
so essentially you'd require >$2.66B to corrupt the ETH quorum on EigenDA. If we add Proof of Custody slashing that would be an increased disincentive.
So when we refer to economic security we often blend in discussing properties we want to protect like this safety threshold.
Hope that gives you a bit of detail on how EigenDA and others work with economic security.