At the moment, a plausible candidate for asymmetric PoReps (with potentially subsecond unsealing) may be one based on a “single(-layer) DRG”. This is a PoRep based on a single layer of a DRG. Notice that in these constructions the data is embedded with the labels through a slightly different approach than the one we use in SDR: not through a simple position-wise XOR with the labels, but with the XOR done before computing the next label (more details below; see figure).
There are really two types of Single DRG constructions. They differ on whether we use all of the layer to encode the data, or if only the tail of the layer is used for that. I will call an approach using all of the tail for data “homogeneous”, while I will call the other approach (head + tail for data)”heterogeneous”:
- Heterogeneous layer. This is described in the “Proofs of Catalytic Space” paper as
- to have security from a provable DRG, this requires high degree of the nodes. This is c log(N) where the constant c is large (approx 1000)
- Homogeneous layer. This is described in Ben Fisch’s “Proofs of Replication” paper as DRG-PoRep. This construction is also described here.
- security (space gap&latency) come directly from the underlying DRG property of the graph. (following notes are from meeting notes) Assume storage is less than e. Then space gap is 1-e. And e+d<1, so d<1-e. So if you want a small space gap, you also get a small latency.
Other possible avenues for asymmetric PoReps with fast unsealing besides Single-DRG:
- Super concentrators (from the original PoS paper)
- Zig Zag (which is insecure at the moment)