### Private Information Retrieval for Ethereum wallets and Portfolio/DeFi frontends

Wallets and DeFi frontends read lots of on-chain data, and we want the privacy of those reads to be preserved with cryptographic guarantees using eg [Private Information Retrieval](https://en.wikipedia.org/wiki/Private_information_retrieval)

### Goals

Add privacy to wallet queries such as:
- Fetch all [cross-chain aggregated] balances of address `0x..`
- Fetch all historical transactions of address `0x..` on chain `y`
- Fetch all active Aave loans of address `0x..`

### Simplifying assumptions

- Small db: ETH/ERC20/NFT balances, including cTokens which reflect DeFi positions
- Omit MPT roots, users trust the integrity of the response, until the [UBT upgrade](https://eips.ethereum.org/EIPS/eip-7864):
- <details>Although, we can (and are) experiment with UBT _today_ as many Ethereum clients have implemented [EIP7864](https://eips.ethereum.org/EIPS/eip-7864) .. there may be a stop-gap solution where a server proves the mpt<>ubt syncronization with each block and attach that prove to queries from helios for example, but that's a different rabbit hole.</details>

### Schemes and implementations:

:arrow_forward: **[See also](https://hackmd.io/@keewoolee/Skvu0BDRle)**

<br>

[](https://github.com/ahenzinger/simplepir)

<br>

**Recent/Active/Notable Implementations**:

- [Plinko](https://eprint.iacr.org/2024/318) "Will need a [~70mb hint on the client side with most active wallets](https://gist.github.com/igor53627/a3e3b258af454309dd6a4941b60cce5b)" -Igor
- [RAID-PIR](https://thomaschneider.de/papers/DHS14.pdf)
- with [onion routing](https://drive.google.com/file/d/13KtaFWpNbjvv4eJIp6PqXHjpPL3xianL/view)
- with [GPU acceleration](https://github.com/encryptogroup/cip-pir?tab=readme-ov-file), [paper](https://encrypto.de/papers/GHPS22.pdf)
- [Rust alt impl](https://github.com/KoffeinFlummi/raidpir?tab=readme-ov-file)
- Hintless [PIR by Google](https://github.com/google/hintless_pir)
- [SimplePIR](https://github.com/ahenzinger/simplepir?tab=readme-ov-file)
- Microsoft [SealPIR](https://github.com/microsoft/sealpir/)
- [Hybrid](https://github.com/KoffeinFlummi/hybridpir?tab=readme-ov-file) with GPU-acc RAID-PIR
- [DP PIR](https://github.com/multiparty/DP-PIR?tab=readme-ov-file), [paper](https://www.usenix.org/system/files/sec22-albab.pdf)
- [FrodoPIR](https://github.com/itzmeanjan/frodoPIR)
- LWE only, no FHE
- "client and server prepare an offline
internal state to be used for making online queries" [[1]](https://eprint.iacr.org/2022/981.pdf)
- _"Each time we added a new breach to the server we’d need clients to update which wasn’t going to work well"_ -feedback from Brave, they tried using it for alterting users of breached passwords
- **[Batched PIR](https://github.com/mhmughees/vectorized_batchpir?tab=readme-ov-file) [[1]](https://eprint.iacr.org/2022/1262)** :eyes:
- <details><br>"Existing BatchPIR schemes have achieved decent computation efficiency but have not been able to improve communication efficiency at all. <br><br>Using vectorized homomorphic encryption, we present the first BatchPIR protocol that is efficient in both computation and communication"</details>
- **[QuarterPIR](https://github.com/wuwuz/QuarterPIR)** :eyes:

### Notes and open questions

- The [Piano](https://www.cs.cmu.edu/~csd-phd-blog/2024/piano-private-information-retrieval/) breakthrough: preprocessing to get around the $\Omega(N$) computation
- **How heavy would pre-processing be in wallet/browser?**
- Pinged Brave Browser to see if they're still actively [exploring PIR](https://github.com/brave-experiments/frodo-pir)
- Families: `Single server` / `Multi-server` / `Hybrid`
- Multi-server is information-theoretic and non-viable performance-wise?
- Paradigms?: `off-line`/`stateful` (~server-heavy) vs `online`/`stateless` (~client-heavy)
- [SeAL](https://github.com/Microsoft/SEAL) is most common dependency
- Levers and tradeoffs, which is best for [our usecase](https://hackmd.io/zNvSA68nQ72JgNUZ2XlP_g?both#Goals):
- Single vs ~~multi-server~~
- FHE or not: what's the catch with FHE-less schemes?
- Pre-processing: how bad?
- Batching
- [Keewoo's slides](https://drive.google.com/file/d/1twAoiKaVuRNbiBUhs_tq-3IB6Jx8UCAf/view)