Data availability
Data availability proofs
On-chain (semi-non-interactive)
Non-interactive
Erasure code
Mustafa Al-Bassam, Alberto Sonnino, and Vitalik Buterin
Mingchao Yu, Saeid Sahraei, Songze Li, Salman Avestimehr (USC), Sreeram Kannan (University of Washington), Pramod Viswanath (UIUC)
FC'20
Background: Erasure coding for data invalidity/unavailability attack in light client/sharding (Al-bassam et al.) Contribution: Propose SPAR protocol
A new hash accumulator named coded Merkle tree (CMT), which encodes every layer of the tree to protect the availability of the entire tree. This way, the Merkle proof of every coded symbol will be available, which will enable every parity equation to be committed and verified alone
https://gyazo.com/f8091adf49b284d61ae7de8bd9819e16
Drawback: Higher sampling cost
The cost of SPAR is about 10∼16 (resp. 2.5∼4) times of 2D-RS under strong (resp. weak) adversaries
STARK proving that the extended Merkle root is correct
Threshold voting
$ n/2BLS signature
The above BLS scheme with erasure codes
Honest-majority availability votes across all shards for just the bandwidth cost of a single shard
Applications
Data availability engine
Mustafa Al-Bassam
Sharding
Daniel Sel, Kaiwen Zhang, Hans-Arno Jacobsen (Technical University of Munich, Middleware Systems Research Group, etc.)
Layer2
Plasma
Tutorials
Vitalik on data availability proofs Video @StarkWare Sessions 2019