Dear Bitcoin Dev Community,

A bit over six months after introducing the P2QRH proposal (now BIP-360), I'm writing to share significant developments and request additional feedback on our post-quantum roadmap, and I'd also like to mention a potential P2TRH post-quantum mitigation strategy.

First, now that there's a BIP number assigned, you can find the update BIP here:

https://github.com/cryptoquick/bips/blob/p2qrh/bip-0360.mediawiki

The revised BIP-360 draft reflects substantial changes since initial publication, particularly regarding algorithm selection. While we originally considered SQIsign, it has 15,000x slower verification compared to ECC [1]. If it takes 1 second to verify a fully ECC block, it would take 4 hours to validate a block filled with SQIsign transactions. This has obvious and concerning DDoS implications.

While it would take a long time to sign many thousands of SQIsign transactions as well, the increased time needed to sign the transactions likely won’t affect the practicality of DDoS attacks-- another concern which has been brought to my attention. As such, I've decided to deprecate SQIsign from the BIP.

It's worth mentioning because it was brought up in the PR, there's a new class of algorithms that support signature aggregation, but they generally result in signatures that are still quite large. Chipmunk and RACCOON are good examples [2], [3]. I do expect that to improve with time. It might be worthwhile to shorten the list by making signature aggregation a requirement, so as not to regress too far from Schnorr signatures. That said, I think those capabilities should be introduced in a separate BIP once they're more mature and worthwhile.

Our current shortlist prioritizes FALCON for its signature aggregation potential, with SPHINCS+ and CRYSTALS-Dilithium as secondary candidates. However, major technical challenges remain, particularly BIP-32 compatibility issues affecting xpub generation in watch-only wallets, as detailed by conduition in another mailing list discussion [4], and also, how we should handle multisig wallets.

Additionally, I think it's worthwhile to restrict BIP-360 to NIST-approved algorithms to maintain FIPS compliance. That's because HSMs such as those provided by Securosys already have support for all three algorithms [5], which is essential for secure deployment of federated L2 treasuries.

Presently, for multisigs, we have a merkle tree configuration defined for encumbering the output with multiple keys. While that's efficient, it's a novel construction. I'm not certain we should proceed with the merkle tree commitment scheme-- it needs more scrutiny. We could use a sort of P2SH approach, just modifying the semantics of OP_CHECKMULTISIG in a witness script to alias to public keys in the attestation. But that could introduce additional overhead in a signature scheme that already uses a lot more space. Without this, however, we do not yet have a way specified to indicate thresholds or a locking script for the attestation, as it is designed to be purposely limited, so as specified it is only capable of n/n multisig. I consider m/n multisigs to be the single largest obvious omission in the spec right now. It definitely needs more thought and I'm open to suggestions. Perhaps two additional bytes at the top level of the SegWit v3 output hash could be provided to indicate PQC signature threshold and total, and those would be hashed and committed to in the output, then provided in a field in the attestation once spent.

While finalizing PQC selections, I've also drafted P2TRH as an interim solution to secure Taproot keypath spends without disabling them, as Matthew Corallo proposes in the aforementioned mailing list thread [4]. The P2TRH approach hashes public keys rather than exposing them directly, particularly benefiting:

- MuSig2 Lightning channel implementations

- FROST-based MPC vaults

- High-value transactions using private pools that don't reveal the block template

For those interested, take a look at the draft BIP for P2TRH here: https://github.com/cryptoquick/bips/blob/p2trh/bip-p2trh.mediawiki

I have my hands full with P2QRH advocacy and development and would prefer to focus on that, but I wanted to introduce P2TRH in case that is attractive as the community's preferred solution-- at least for Taproot quantum security. The tradeoff is that it adds 8.25 vB of overhead per input, and key tweaking might have slightly less utility for some applications, and it also doesn't protect against short exposure quantum attacks as defined in BIP-360.

Returning to P2QRH and what's needed to push it across the finish line...

I still need to finish the test vectors. I'm implementing these using a fork of rust-bitcoin and modeling them after Steven Roose's work on BIP-346. I've been told that's not a blocker for merging the draft, but if it isn't merged by the time I'm finished, hopefully that will provide some additional impetus behind it.

One concern Murch brought up is that introducing four new algorithms into the network was too many-- adding too much complexity to the network and to wallets and other applications-- and I agree. 

Hopefully this is addressed to some degree by removing SQIsign (especially in its current state lacking implementation maturity), and will help push the BIP below a certain complexity threshold, making it somewhat easier to review.

 

I think it's still important to include multiple signature algorithm options for users to select their desired level of security. It's not 100% certain that all of these algorithms will remain quantum resistant for all time, so redundancy here is… key.

Another concern is that NIST level V is overkill. I have less conviction on this since secp256k1 technically has 128 bits of security due to Pollard's rho attacks. But if the intention was for 256 bits of security, should level V security be the default? It's difficult for me to say. Perhaps both level V and level I implementations could be included, but this would be a deviation from the BIP as presently specified, which defaults to level V security. The disadvantage of including level I support for each algorithm is that it essentially doubles the complexity of libbitcoinpqc.

Ultimately, I hope the default of NIST V and selection of 3 mature NIST-approved algorithms demonstrate a focused, polished, and conservative proposal.

At this point, the major call to action I would like to highlight is simply the need for more feedback from the community. Please review and provide feedback here: https://github.com/bitcoin/bips/pull/1670

I look forward to feedback and opinions on P2QRH and P2TRH.

P.S. I'll be advocating for BIP-360 at OP_NEXT in VA, btc++ in Austin, Consensus in Toronto, and BTC 25 in Las Vegas, and later this year, TABConf in Atlanta.

[1] https://pqshield.github.io/nist-sigs-zoo

[2] https://eprint.iacr.org/2023/1820.pdf

[3] https://eprint.iacr.org/2024/1291.pdf

[4] https://groups.google.com/g/bitcoindev/c/8O857bRSVV8/m/7uu4dZNgAwAJ

[5] https://docs.securosys.com/tsb/Tutorials/Post-Quantum-Cryptography/pqc-release-overview

--
You received this message because you are subscribed to the Google Groups "Bitcoin Development Mailing List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bitcoindev+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/8797807d-e017-44e2-b419-803291779007n%40googlegroups.com.