Thread by @RhysLindmark

1/ Thread from the @SimonsInstitute workshop on:

Blockchain in Society: Applications, Economics, Law, and Ethics.

Schedule and livestream here: simons.berkeley.edu/workshops/sche…

1a/ First, Kobbi Nissim on "Predicate Signaling Out". (PSO)

AFAIU, it is a technical definition of privacy that formalizes the legal requirements of GDPR. (Asks the question—when could an adversary "signal out" a single person/row from a database?)

arxiv.org/abs/1904.06009

1b/ I didn't really understand the math of PSO, but some other learnings:

-Differential privacy (DP) and k-anonymity are defined mathematically. Contextual Integrity and Robot Lawyers are defined with formal logic.
-Nissim tried to create a math def of FERPA. It ended up ~= DP.

1c/ Intuitively, k-anonymity should conform with PSO. (k-anon creates k rows that are anon w.r.t. each other. PSO is all about singling out *one* row.)

But, surprisingly, k-anon does not conform with PSO! (Why? Maths that I didn't understand.)

2a/ @BKCHarvard's @yaoeo now begins a talk on "Regulation through Governance".

First, Prim explains on-chain vs. off-chain governance.

- On-chain is enforcement BY the infrastructure. Enforcement of rules.
- Off-chain is enforcement OF the infrastructure. Creation of rules.

2b/ @yaoeo uses the DAO and the Parity bug to highlight @aantonop's idea of "can't and won't" (w.r.t. immutability).

Crucial ideas:
- Ability leads to liability.
- Formalization (of process) leads to ability.
- Therefore, keep things informal.

@yaoeo @aantonop 3/ I don't have deep notes on @mysteryfigure's talk (on Game Theory).

But I do love this graph (that shows how mining pools have increased as a % of total hash rate).

nber.org/papers/w25592.…

@yaoeo @aantonop @mysteryfigure 4a/ @elibensasson from @starkwareltd gives a talk on "Inclusive Accountability at Scale via Proofs".

See @riabhutoria's: medium.com/circle-researc…

4b/ @EliBenSasson Key idea is: What are the best ways to get *computational integrity*? (Executing a program and reporting its output in a reliable way.)

4 ways:
- Naive Replay
- Delegated Accountability
- Trusted Execution Environments (SGX, etc.)
- ZKPs

4c/ @EliBenSasson

Each has tradeoff space around privacy, scalability, trust, and attack vectors.

Right now, 99% of computational integrity is done with Naive Replay or Delegated Accountability. But Eli is (obviously) excited by using more ZKPs.

@EliBenSasson 4d/ Good addition from @josephbonneau:

To add a 5th option for CI: fraud-proofs (e.g. @Truebitprotocol @offchainlabs).

With these, the trust is in "economics". (The game theoretical equilibrium of the incentives.)

@EliBenSasson @josephbonneau @Truebitprotocol @OffchainLabs 4e/ imo, @EliBenSasson does a great job of responding to questions.

1. He listens to the question/comment
2. Then reflects it. "If I understand correctly, you're saying..."
3. *Then* responds.

Almost no one does #2. But it's crucial!

@EliBenSasson @josephbonneau @Truebitprotocol @OffchainLabs 4f/ @EliBenSasson moves to accountability.

Traditionally, we've had "delegated accountability": use a trusted third party to verify the system.

With blockchains, we have "inclusive accountability": *anyone* can verify the system.

But this sacrifices privacy and scalability.

@EliBenSasson @josephbonneau @Truebitprotocol @OffchainLabs 4g/ And the key idea is that ZKPs (zkSTARKs) can add privacy and scalability, while still having inclusive accountability.

(The best of both worlds!)

@EliBenSasson @josephbonneau @Truebitprotocol @OffchainLabs 5a/ *Super* interesting discussion by @angela_walch on "Miners as Blockchain Intermediaries."

tl;dr: claiming that blockchain doesn't have intermediaries is wrong. Miners are intermediaries.

5b/ @angela_walch likes to make these controversial/brave claims. :)

e.g. The crypto world has a strong norm of:

"Crypto devs != fiduciaries"

and

"Crypto miners != intermediaries"

(Angela has pushed back on both.)

5c/@angela_walch on how miners are intermediaries

-Only miners add new tx, and can change time/order
-Can front run (@phildaian's Flash Boys 2.0)
-Censor (Zcash pool @f2pool_official wasn't adding shielded tx b/c they didn't add that code)
-Rewind (@cz_binance's casual proposal)

5d/ @angela_walch with more on miners as intermediaries:

- How miners incentives change w/ L2 (see @paddypisa's homepages.cs.ncl.ac.uk/patrick.mccorr…)
- Also see @_prestwich's overview here: blog.keep.network/miners-arent-y…

5e/ @angela_walch legal decisions are made w/ miners as *not* intermediaries:

- FinCEN doesn't think they're a money transmitter
- NY Bitlicense didn't categorize miners as such

And with miners *as* intermediaries:
- OFAC blacklisted Bitcoin addresses

5f/ @angela_walch w/ great convo at this point w/ @interfluidity @valkenburgh

- Should ISPs count as intermediaries? (Answer seems to be that they would get in trouble for OFAC addresses.)
- From FinCEN POV, does "adding tx" count as "accepting and transmitting" $$?

5g/ @angela_walch with open research qs:

How differentiate intermediaries?
- Intermittent intermediaries. (Miners can only censor tx *some* of the time.)
- Multiple intermediaries. (Miners as part of governance process to determine forks.)
- Types (mining pools vs. cloud etc.)

5h/ And more research questions from @angela_walch:

- To achieve outcomes, when should we trust game theoretical incentives vs. regulation?
- @avivz78 Should miners be regulated as critical infrastructure? (From cybersecurity POV.)

@angela_walch @Avivz78 6a/ @josephbonneau on Blockchains and Publicly-Verifiable Lotteries.

First up, this interesting (non-formal) consensus trilemma. I've never seen this before. It's a pretty way to conceptualize classic BFT vs. Round-Robin vs. PoW/PoS.

@angela_walch @Avivz78 @josephbonneau 6b/ @josephbonneau w/ a really interesting section on different options for "getting non-cryptographic randomness":

- Natural. (How hot is it?) Issue is different sensors.
- Trusted 3rd-party. (Random.org) Issue is trust. Just a company.

6c/@josephbonneau

-Viewable physical randomness. (Pulling lottery balls.) But can be attacked. (Lead balls.)
-Financial data. In US, lotteries were illegal until 1950. So organized crime did it. They used NSYE digits for randomness. (But can be attacked by messing w/ markets.)

@josephbonneau 6d/ @josephbonneau now looking at cryptographic randomness protocols. Good overview of options.

6e/ @josephbonneau with a *super clear* overview of the different kinds of VDFs.

a) You can take away V to get DF (just repeated hashing).
b) But if you want V, you can naively just SNARK the DF. (But super expensive.)

How do V better?

6f/@josephbonneau w/ 3 options for V, all of which were co-invented June 2018: eprint #'s 601 623, 627. (!)

-Still snarky: eprint.iacr.org/2018/601.pdf
-Variants on repeated squaring, which is log(n) efficient to V by binary searching. eprint.iacr.org/2018/623.pdf eprint.iacr.org/2018/627.pdf

@josephbonneau 7/ @valkenburgh w/ a great synthesis of securities law from 1st principles.

- Regulation is NECESSARY if asset is bought for investment, not utility. (¬ bottom row)
- Regulation is JUST if we can find a "least cost avoider". A single issuer not a huge network. (¬ left column)