3 ms·

I've been wondering about this too. I don't see how the incentives are setup that the Association would want to move towards an entirely permissionless network.

by sigmar 7y ago

I've been wondering about this too. I don't see how the incentives are setup that the Association would want to move towards an entirely permissionless network. Especially if Libra gets heavily adopted at 100 nodes (as they don't need to improve trust at that point and letting more companies into the association will decrease transaction throughput).

nybble41 7y ago
There are also some technical barriers on the path to a permissionless network. The consensus algorithm Libra is using currently requires full connectivity between validators and is only robust so long as 2/3 of the validators can be trusted. Much has been made of the risk of a "51% attack" against Bitcoin, but one need only control 1/3 of the validator nodes to disrupt Libra, which is a far lower threshold than 51% of computation capacity. Your standard botnet isn't much of a threat to Bitcoin since the participating devices aren't optimized for mining, but Libra would be an entirely different matter.
- woah 7y ago
  The 1/3rd / 2/3rd pattern in a lot of blockchain BFT protocols is really elegant. If 1/3 of the voting power of a chain wants it to stop, then it probably shouldn't be running. It takes 2/3rds of the voting power for the validators to actually steal anyone's money or alter history. Your botnet example is a non-sequiter. Since power in BFT protocols does not stem from the wasting of electricity, botnets have nothing to do with any possible attacks on a BFT system.
  - nybble41 7y ago
    On the contrary, botnets are far more of a problem in systems designed to make it cheap to run a validator node. This creates an obvious opportunity for a Sybil attack: spawn a few thousand nodes on other people's insecure IoT devices and you immediately control far more than the requisite 1/3 or 2/3 of the network. It's a lot harder to control 51% of the hash rate for a PoW design like Bitcoin which effectively requires specialized hardware to mine competitively. I'm assuming they implement countermeasures against simply running a bunch of VMs or containers masquerading as a large number of independent nodes, for example by looking for multiple connections from the same IP addresses or subnet. Otherwise botnets would indeed be unnecessary to carry out the attack—anyone could do it from home with a modest PC.

by sigmar 7y ago

nybble41 7y ago

There are also some technical barriers on the path to a permissionless network. The consensus algorithm Libra is using currently requires full connectivity between validators and is only robust so long as 2/3 of the validators can be trusted. Much has been made of the risk of a "51% attack" against Bitcoin, but one need only control 1/3 of the validator nodes to disrupt Libra, which is a far lower threshold than 51% of computation capacity. Your standard botnet isn't much of a threat to Bitcoin since the participating devices aren't optimized for mining, but Libra would be an entirely different matter.

woah 7y ago
The 1/3rd / 2/3rd pattern in a lot of blockchain BFT protocols is really elegant. If 1/3 of the voting power of a chain wants it to stop, then it probably shouldn't be running. It takes 2/3rds of the voting power for the validators to actually steal anyone's money or alter history. Your botnet example is a non-sequiter. Since power in BFT protocols does not stem from the wasting of electricity, botnets have nothing to do with any possible attacks on a BFT system.
- nybble41 7y ago
  On the contrary, botnets are far more of a problem in systems designed to make it cheap to run a validator node. This creates an obvious opportunity for a Sybil attack: spawn a few thousand nodes on other people's insecure IoT devices and you immediately control far more than the requisite 1/3 or 2/3 of the network. It's a lot harder to control 51% of the hash rate for a PoW design like Bitcoin which effectively requires specialized hardware to mine competitively. I'm assuming they implement countermeasures against simply running a bunch of VMs or containers masquerading as a large number of independent nodes, for example by looking for multiple connections from the same IP addresses or subnet. Otherwise botnets would indeed be unnecessary to carry out the attack—anyone could do it from home with a modest PC.