Ideas and Theory
Proof of Space research is relatively new and suffers from a lack of formalization and not very formal definitions. Ideally, we aim to decrease the difficulty of researching Proof of Space, review old protocols, state better definitions and security requirements and design new schemes with better features.
Desired Features and Goals
- Must have: a security margin of 10x at roughly the same costs for storage provider as now; constructions that help against security erosion
- Must have: at most 10% higher costs, similar or at most 10% worst overall costs than current PoRep
- Nice to have:
- faster retrievability, we want to be able to at least improve time to first byte
- Nice to have:
- tunability: understanding what knobs need can be changed/how in the PoS to achieve what other outputs in the network
- NB: a non-PoS problem related to tunability is that changes should be easily deployable (e.g. through a universal setup). PoS-related efforts that help with tunability may not be as effective without that.
Problem 1: Erosion
From a security perspective
- better tuning (intuition: static, upgrade done once or unfrequently)
- option: windowpost more frequent (but this is complicated, not a solution in the long term)
- option: find stronger parameters keeping the frequency
- for SDR there are not many options
- maybe starting from a primitive with better “bounds we are aware of”
- Better EE is in this item in this sense.
Problem 2: Tunability
The property of being easily upgradable often
- hard to define but here is an example
- suppose you want to upgrade frequently and you do need to change graph, SNARK setup size, etc.
- this is non-tunable
- something “tunable” is something that does not have this issue
Problem 3: Retrieval
- two problems now:
- retrieval is slow (← this is the one we can work in a PoS project of course)
- Saturn may mitigate the priority of this requirement somewhat but the bottleneck remains in the retrieval-from-SP
- there are no incentives to do retrieval (hard to solve cryptographically; we ignore this)