Difference between revisions of "Time Attack"
wiki_crypto>Zeb.dyor (Created page with "* From [https://ethresear.ch/t/time-attacks-and-security-models/6936 this post] (13-2-2020) on Eth Research, which goes into it a lot more: ''"Many [[Proof-...") |
m (1 revision imported) |
Latest revision as of 09:01, 23 January 2022
"Many Proof-of-stake designs assume validator clocks are roughly synchronized. Often the clock sync property is taken for granted, but, in reality, clocks should be synchronized with some protocol. Thus, such clock sync protocol becomes a target for an attacker.
Particularly, in beacon chain protocol, if a validator’s clock disparity is more than one epoch, then its attestations won’t be included in a chain by correct processes. Due to inactivity lick, the validator will be penalized for not participating in the protocol.
Time attack is attractive because its cost can be much lower than the cost of validator deposits.
Assuming, there are 300K validators and 10K nodes, there are 30 validator per node on average. To perform 51% attack directly, an attacker should own 150K validators, which would cost 4.8M ETH (about $1.3B at the time of writing).
However, if many validating nodes use NTP with insecure setups, they can be attacked with much lower cost (see, for example, one, two, three)."