zkDaily - 2026-08-18

零知识证明zkDaily

ZKP Frontier Tracker 🎯

Tue

08.18

2026

TeleZK-FL: enabling trustless and verifiable remote patient monitoring via quantized zero-knowledge federated learning

Paper

https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2026.1821129/full

Federated LearningHalo2LUT

Jayaraman et al. proposed TeleZK-FL, combining INT8 post-training quantization with LUT arguments to enable on-device zero-knowledge proving for verifiable federated learning, achieving ~25x faster proofs.

NotesNotes要点

TeleZK-FL quantizes gradients to INT8, cutting communication payload by 75%.
Halo2 LUT arguments decouple constraint cost from arithmetic bit-width.
Full-model proof generation takes only 84 ms on Raspberry Pi 4.
AUC drops only 0.1% on CheXpert and 0.3% on PTB-XL.
Security relies on KZG commitments, classically secure but not post-quantum.
Defends against large-norm gradient poisoning but not low-norm backdoors.

Collected by @icerdesign

零知识证明zkDaily

Q&A Deep Dive 💬

Tue

08.18

2026

What problem does TeleZK-FL solve?

It enables resource-constrained medical edge devices to generate zero-knowledge proofs without trusted servers, achieving verifiable federated learning.

How does it speed up proof generation?

By replacing arithmetic constraints with LUT arguments, cost is decoupled from bit-width, achieving ~25x speedup over standard zk-SNARKs.

What poisoning defense is guaranteed?

The circuit enforces an L2 norm range check, rejecting large-norm updates, but it does not stop low-norm backdoors; robust aggregation is complementary.

Prompted by @icerdesign