Project Summary
We propose to extend an existing machine-checked Pallas arithmetic MVP into reproducible, source-linked scalar-decomposition proofs for Pallas and Vesta. The project will connect the supported Rust scalar representations and serialization path to the arithmetic guarantees, then publish the proofs, replay tooling and precise coverage documentation.
Project Description
We will develop reproducible, source-linked proofs of GLV scalar decomposition for Pallas and Vesta, strengthening assurance for a cryptographic component used in Zcash wallet scanning.
The project builds on an existing machine-checked Pallas MVP. Grant-funded work will connect the supported Rust scalar representation, serialization and limb conversion to the arithmetic proofs, extend coverage to Vesta, and establish that the resulting components remain within the required bounds and reconstruct the original scalar.
Deliverables include proof sources, pinned toolchains, replay scripts, validation controls and documentation of coverage and assumptions, followed by 90 days of bounded maintenance. The work covers the portable implementation, excluding assembly, constant-time behavior and whole-wallet security.
MVP: GitHub - Crackdevs/zcash-glv-scalar-proof · GitHub
Proposed Problem
GLV scalar decomposition produces smaller signed components used in efficient elliptic-curve multiplication. Correctness requires those components to reconstruct the original scalar modulo its order and remain within the bounds expected by subsequent algorithms.
Pasta’s issue #174 explicitly identifies an all-scalar decomposition proof as an outstanding part of GLV formalization. Orchard’s released batched trial-decryption implementation uses GLV, giving this work a concrete connection to Zcash wallet scanning.
The existing MVP establishes the arithmetic guarantees for canonical Pallas limb inputs. It does not yet prove the complete connection from the supported Rust scalar representation, through serialization and limb conversion, to that arithmetic result.
This project closes that specific assurance gap and it does not claim that an exploit has been discovered or that all GLV scalar multiplication will be verified.
Sources:
zcash/pasta_curves#174
Proposed Solution
We will connect the portable Rust scalar-representation, serialization and limb-conversion path to machine-checked Lean proofs of GLV decomposition for Pallas and Vesta. This will establish that every valid supported input produces bounded components that reconstruct the original scalar modulo its order.
The work will extend the existing Pallas arithmetic MVP, prove the missing representation conversions and complete the corresponding Vesta proof. A pinned source and toolchain, reproducible replay scripts, negative controls and explicit documentation of assumptions will allow developers and reviewers to check the results independently.
An initial feasibility milestone will confirm the source-to-proof approach before full implementation proceeds. The final package will be publicly available without requiring a wallet deployment or upstream merge.
Solution Format
An open-source formal-verification package containing Lean proofs for the supported Pallas and Vesta GLV decomposition paths, pinned Rust source and toolchain manifests, extraction and replay scripts, validation controls, and documentation of coverage and assumptions. Delivery includes a versioned release and 90 days of bounded reproducibility support.
Dependencies
The project depends on the pinned pasta_curves source, Lean, Aeneas/Charon, Mathlib and Crackdevs’ technical team. Versions and required local changes will be documented. Milestone validation will follow the process agreed with ZCG and the intended users or their representatives. Delivery does not depend on an upstream merge or wallet deployment.
Technical Approach
The baseline is pasta_curves 0.6.1 at commit 674bf82910d19fefdcad323d7dd05e8c77baa0bd, with GLV enabled and assembly disabled. We will establish a reproducible translation of the concrete Rust serialization and limb-conversion path, define the supported scalar-representation invariant, and prove the required Montgomery-to-canonical and byte-to-limb correspondence.
These results will be composed with the existing Pallas arithmetic proof, with corresponding proofs completed for Vesta. Acceptance requires successful computation, component magnitudes strictly below 2^127, and modular reconstruction for every valid supported input. Source correspondence, trusted tooling and axioms will be documented. Clean replay and meaningful arithmetic mutation controls will validate the package; source hashes will detect changes, not establish semantic equivalence.
Upstream Merge Opportunities
The primary deliverable will remain in the standalone Crackdevs proof repository. The project will use a pinned copy of zcash/pasta_curves for extraction and verification, without requiring changes to deployed Zcash software.
Reusable proofs, extraction improvements or documentation may be offered to pasta_curves or relevant formalization projects. Any upstream contribution will follow the receiving project’s guidelines and review process.
Open Source License
MIT
Hardware/Software Costs (USD)
0
Hardware/Software Justification
None
Service Costs (USD)
0
Service Costs Justification
NIL
Compensation Costs (USD)
$13,200
Compensation Costs Justification
- Proof implementation and source correspondence: 80 hours × $125 = $10,000.
- Release packaging and documentation: 20 hours × $100 = $2,000.
- Bounded maintenance: 12 hours × $100 = $1,200.
These are planning estimates. Payments are tied to accepted milestone deliverables, not hours alone. Completed MVP work is excluded. Milestone one will validate the remaining obligations and delivery estimate before the larger implementation stage proceeds.
Total Budget (USD)
$13,200
Previous Funding
No
Previous Funding Details
No response
Other Funding Sources
No
Other Funding Sources Details
No response
Implementation Risks
The principal risk is that connecting the actual Rust representation and serialization path to the Lean proofs requires more extraction or specialization work than estimated. Milestone one will establish the approach, trust assumptions and remaining proof obligations. If the work exceeds the agreed scope, subsequent milestones will pause pending agreement. Source and toolchain versions will be pinned to limit uncontrolled changes.
Potential Side Effects
The principal risk is false confidence from interpreting a limited proof as comprehensive security assurance. Documentation will distinguish proved properties, assumptions and excluded behavior. The project does not change consensus, custody funds or deploy a wallet service. Assembly, compiler correctness, constant-time behavior, full scalar multiplication and whole-wallet security remain outside scope.
Success Metrics
- Source-linked decomposition proofs for both supported curve instantiations, with explicit representation invariants.
- Proofs of successful computation, component bounds and modular reconstruction.
- Successful clean replay using pinned source and toolchains.
- Published theorem, axiom and coverage inventories, with no unfinished target proofs or assumed target arithmetic conclusions.
- Meaningful negative controls, a versioned release and a maintenance closeout report.
Startup Funding (USD)
0
Startup Funding Justification
None
Milestone Details
Milestone 1: Establish the source connection
Amount (USD): 2500
Expected Completion Date: 2026-11-14
User Story:
As a reviewer, I want to trace the proof definitions to the selected Rust source so that I can assess their coverage and assumptions.
Deliverables:
- Pinned source and toolchain manifests.
- Reproducible extraction instructions and compiling translated definitions.
- A typed composition interface and supported scalar-representation invariant.
- A source-to-proof coverage map, documented trust assumptions and remaining-work estimate.
Acceptance Criteria:
The agreed source path and interface compile, their correspondence is documented, and the remaining proof obligations are identified. If the approach is infeasible within scope, later milestones pause. A blocker report is payable only if explicitly agreed and accepted as an alternative deliverable.
Milestone 2: Complete the Pallas and Vesta proofs
Amount (USD): 7500
Expected Completion Date: 2026-11-28
User Story:
As a library developer, I want decomposition guarantees for every valid supported scalar input, rather than relying only on sampled tests.
Deliverables:
- Serialization and limb-conversion correspondence proofs.
- Concrete Pallas and Vesta proof compositions.
- Documentation of specialization, source correspondence and build-profile restrictions.
Acceptance Criteria:
For every input satisfying the stated representation invariant, the linked computation succeeds, both signed component magnitudes are strictly below 2^127, and their lambda-weighted sum reconstructs the original scalar modulo its order. These arithmetic conclusions are proved rather than assumed. Work proceeds following milestone-one acceptance.
Milestone 3: Publish the reproducible release
Amount (USD): 2000
Expected Completion Date: 2026-12-07
User Story:
As a reviewer, I want to replay the proofs and inspect their assumptions using a documented, reproducible environment.
Deliverables:
- A versioned release with clean-environment replay instructions and logs.
- Theorem and axiom inventories, plus explicit coverage documentation.
- Meaningful arithmetic mutation controls and source-drift checks.
- Internal technical-review notes and responses to milestone-validation findings.
- Licensing, attribution and CONTRIBUTING.md.
Acceptance Criteria:
Successful clean replay, no unfinished target proofs, explicit trusted assumptions, meaningful negative controls and resolution of acceptance-blocking findings within the agreed scope. Validation follows the agreed intended-user or representative process. Source-drift checks identify changed inputs rather than prove their correctness. Milestone acceptance is not presented as independent audit certification.
Milestone 4: Maintenance and closeout
Amount (USD): 1200
Expected Completion Date: 2027-03-07
User Story:
As a proof-package user, I want reproducibility problems to receive a documented response during the support period.
Deliverables:
- A 90-day support period following the milestone-three release, covering up to 12 engineering hours of pinned-release reproducibility support and small compatible corrections.
- A public support log.
- A closeout report with final replay results.
Acceptance Criteria:
Completion of the agreed support window and publication of the support log and closeout evidence. Major dependency migrations, new algorithms and additional backends are excluded. If the release date changes, maintenance will end 90 days after the actual release.
Supporting Documents
-
Public MVP repository: GitHub - Crackdevs/zcash-glv-scalar-proof · GitHub
Contains the existing proof sources, replay tooling, recorded results and documented limitations. -
Scope and findings: zcash-glv-scalar-proof/FINDINGS.md at main · Crackdevs/zcash-glv-scalar-proof · GitHub
-
Upstream problem statement: Formalize the GLV scalar multiplication · Issue #174 · zcash/pasta_curves · GitHub