1.15 Million Auth/Sec
Full Cryptographic Stack
FHE biometric matching + Dilithium signatures + ZK proofs. Every authentication is encrypted, signed, and zero-knowledge proven. One API call. Production ARM hardware. Measured over 60 seconds sustained.
This case study measures the cryptographic lifecycle whose portable evidence H33-74 produces — on AWS Graviton4 ARM hardware. Why this exists: post-quantum security is often assumed to be too slow for production, so this run reports reproducible throughput for the full stack. All figures below are from a single measured run; see Benchmarks for methodology and reproduction.
Per-Batch Pipeline
Each batch processes 32 users in a single FHE operation. One Dilithium signature attests the entire batch.
Sustained Throughput
60 seconds sustained, optimized allocator. Every batch: FHE biometric + Dilithium batch attestation.
1,114K
1,128K
1,170K
1,193K
1,141K
1,205K
Component Latencies
Isolated measurements on production ARM hardware (aarch64). No contention.
ZK Prover Performance
STARK zero-knowledge proofs with SHA3-256 hash commitments. Native H33 implementation.
MSM parallelization: thread-local buckets per window with merge. IPA recursive folding thresholded at n ≥ 64 to avoid overhead on small recursion.
End-to-End Latency Distribution
Dynamic batching with 2ms max wait. Measured at increasing request rates.
Post-Quantum Security Stack
Every component is quantum-resistant. No classical-only primitives anywhere in the pipeline.
Benchmark Configuration
Production ARM hardware. No GPUs. No FPGAs. Pure CPU.
FHE Parameters
Test Configuration
Biometric authentications per second. Each one encrypted with FHE, attested with Dilithium, and provable with zero-knowledge proofs. Fully post-quantum. Production ARM hardware. This is not a projection. This is a 60-second sustained measurement.
Case Study FAQ
How to read these numbers, and which mechanism owns each capability.
What does this case study measure?
A single measured, reproducible run of the full post-quantum stack — FHE biometric matching, Dilithium (ML-DSA-65) attestation, and ZK-STARK proofs — on AWS Graviton4. It exercises the cryptographic lifecycle whose portable evidence H33-74 produces. Reproduction methodology lives at Benchmarks.
How is biometric matching done without exposing the biometric?
FHE computes on encrypted data: BFV runs the encrypted inner product and a 3-of-5 threshold decrypt reveals only the match result — the biometric template is never decrypted on the server.
What do the Dilithium signature and ZK proof each contribute?
The Dilithium (ML-DSA-65) batch attestation binds each result into the portable receipt H33-74 produces, and ZK proves without revealing the underlying inputs. Whether such a receipt is trustworthy is a separate question answered by Verification, which renders the independent verdict.
Are these numbers projections?
No — they are from a 60-second sustained measurement on production ARM hardware, not a projection. Figures are reproducible; see Benchmarks for the configuration and method.
See also — canonical owners: H33-74 produces portable evidence · FHE computes on encrypted data · ZK proves without revealing · Verification renders the verdict · Benchmarks holds the methodology.