Grant Application - Post-quantum readiness inventory for the Zcash stack

A note for reviewers, on Epoch and Tachyon

Shielded Labs announced Epoch
on 1 October — production-ready, high-assurance post-quantum cryptography for Zcash, with formal verification,
working closely with Daira-Emma Hopwood and Dev Ojha and with Sean Bowe, Tal Derei and the rest of the Project
Tachyon team, and targeting a production-ready cryptographic implementation by the end of 2027. Zcash also has a
staged plan already moving: ZIP 2005 quantum recoverability on testnet, ML-KEM alongside Tachyon, then a fully
post-quantum pool.
“Epoch is already doing this” is the first reasonable objection to this application, so I would rather answer it
myself than have a reviewer raise it.

Epoch and Tachyon build the new cryptography. This proposal inventories the classical cryptography that is
deployed in the repositories today and will have to be migrated to it. Those are different objects. Of the 112
inventory rows across the six published audits, 40 are Shor-breakable:

Repository Shor-breakable rows What they are
librustzcash 19 15 × ECDSA secp256k1 (transparent addresses), 4 × BLS12-381
zebra 14 5 × Ed25519, 4 × X25519, 2 × BLS12-381, 2 × Groth16, 1 × ECDSA secp256k1
lightwalletd 3 RSA-2048 — certificate generation in common/generatecerts.go and the self-signed docker/cert.pem
zcash-android-wallet-sdk 2 RSA-2048 quoted in responsible_disclosure.md
zcash-swift-wallet-sdk 2 RSA-2048 quoted in responsible_disclosure.md

Two things in that table I would rather point out than be asked about. The last four rows are a PGP key quoted in
a disclosure policy: that is inventory, not exposure, and I do not count it as risk. And the two Groth16 rows in
zebra are the one place where “Epoch is already doing this” is literally true — replacing the proving system is
Epoch’s and Tachyon’s work, not mine. What remains after both caveats is transparent-address and transport
cryptography: ECDSA secp256k1, Ed25519, X25519, RSA certificate keys, spread across call sites in five
repositories. A new signature scheme does not find those call sites. Somebody has to, once, and then keep the list
current as the code moves.

It has already been pointed out in Is Zcash Post-Quantum?
that most of Zcash’s cryptography is symmetric hashing, which Grover weakens rather than Shor breaking. The point
of an inventory is to settle that by counting rather than arguing, so here is the count for these six
repositories. Of the 112 rows, 52 are crypto libraries declared in manifests and lock files — secp256k1,
rsa, ecdsa, p256, p384, ring, rustls, the dalek curves, Bouncy Castle — which I record as inventory
and deliberately do not score, because the audits do not establish that each one is reachable. That leaves 60 rows
of actual algorithm use, and the split is 40 Shor-breakable against 20 that are a Grover question (12 × SHA-2,
2 × BLAKE2, 1 × Keccak-256, 2 × ChaCha20-Poly1305) or a classical weakness (2 × RIPEMD-160, 1 × SHA-1). Two
thirds, not a minority — and per row, with the file and line.

One of those twenty is worth a look on its own, as an example of what the inventory is for.
zcash-android-wallet-sdk validates the Sapling parameters it downloads against two hard-coded SHA-1 digests,
SPEND_PARAM_FILE_SHA1_HASH and OUTPUT_PARAM_FILE_SHA1_HASH in sdk-lib/.../SaplingParamTool.kt:66,71, and the
comment at line 273 says plainly that the check is there because the file may arrive “from a malicious server”.
This is not exploitable: it rests on second-preimage resistance, which SHA-1 still has. But Grover takes that
from 160 bits to roughly 80, which is below any margin I would accept for parameters with that lifetime. Note also
what my tooling did and did not do: it flagged the SHA-1 call in backend-lib/.../FileExt.kt:50, which is where
the algorithm is — not SaplingParamTool.kt, which is where the two digests and the threat model are. The scanner
produces the row; connecting it to the threat model is my work, and the report distinguishes the two.

So I am not proposing an alternative to Epoch. I am proposing the input that Epoch and the wallet teams would
otherwise have to produce for themselves, and I would rather be weighed on that basis than on the broader
post-quantum framing. If the committee’s view is that this belongs inside Epoch or ECC rather than as an
independent grant, I think that is a fair conclusion — the work is the same either way, and the six audits already
published are free to whoever wants them.