惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Google DeepMind News
Google DeepMind News
aimingoo的专栏
aimingoo的专栏
爱范儿
爱范儿
D
Docker
I
InfoQ
Microsoft Security Blog
Microsoft Security Blog
G
Google Developers Blog
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
Vercel News
Vercel News
H
Hackread – Cybersecurity News, Data Breaches, AI and More
T
Tailwind CSS Blog
D
DataBreaches.Net
月光博客
月光博客
N
Netflix TechBlog - Medium
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
V
Visual Studio Blog
MyScale Blog
MyScale Blog
B
Blog
阮一峰的网络日志
阮一峰的网络日志
L
LangChain Blog
Recent Announcements
Recent Announcements
Microsoft Azure Blog
Microsoft Azure Blog
WordPress大学
WordPress大学

Hacker News: Show HN

PurrrrrFocus: Pomodoro Timer App - App Store Workflow Engine — Multi-Step Orchestration for Bun RapidPhoto: Pro Photo Editor App - App Store GitHub - DheerG/swarms: Achieve extraordinary results with claude code across a variety of tasks SPICE simulation → oscilloscope → verification with Claude Code — Lucas Gerads Show HN: VCoding – A 5 MB native Windows IDE with no dynamic dependencies Show HN: LLMs don't hallucinate because they're bad at math, it's the format GitHub - Agent-FM/agentfm-core: AgentFM is a peer-to-peer network that turns everyday computers into a decentralized AI supercomputer. AgentFM lets you run massive AI workloads directly across a global mesh of idle CPUs and GPUs. Show HN: Tracking Top US Science Olympiad Alumni over Last 25 Years GitHub - Potarix/agent-hub: One place to talk to all your agents Show HN: Runtime security for AI agents(injection,tool abuse, data exfiltration) GitHub - dubeyKartikay/lazyspotify: Terminal Spotify client for macOS and Linux GitHub - the-banana-tool/king-louie: Easy to use GUI Personal AI Assistant. Win/Linux/Mac. Show HN I made my vacation rental bookable by AI agents–no Airbnb, 0% commission GitHub - basteez/jsf-autoreload: maven plugin to enable hot reload on jsf projects uvm32/hosts/host-gdbstub at main · ringtailsoftware/uvm32 GitHub - labsai/EDDI: Config-driven engine that turns JSON into production-grade AI agents. Multi-agent orchestration, 12+ LLM providers, MCP/A2A protocols, RAG, persistent memory, and enterprise compliance (EU AI Act, GDPR, HIPAA). Built on Quarkus. GitHub - glitchnsec/fortyone-oss: AI Executive Assistant Platform Quickstart | Alien GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. GitHub - ocrbase-hq/ocrbase: 📄 PDF/IMG ->.MD/JSON Document OCR API for PaddleOCR and GLMOCR. Self-hostable. GitHub - impactjo/home-memory: MCP server that lets your AI assistant remember everything about your home. GitHub - Sets88/dbcls: DbCls is a powerful terminal database client that supports various databases GitHub - neptun2000/heor-agent-mcp GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh RollQuation: Math Puzzles - Apps on Google Play GitHub - dropbox/witchcraft Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis GitHub - opentalon/opentalon: OpenTalon is an open-source platform built from the ground up in Go as a robust alternative to OpenClaw LinkedIn™ 职位抓取工具 - Chrome 应用商店
GitHub - unprovable/OrchidMantis: Orchid Mantis — standal...
unprovable · 2026-05-16 · via Hacker News: Show HN

A Framework for ZKPoX — Zero-Knowledge Proofs of Exploit

Orchid Mantis

Status: experimental (v0.1). Bundle format, predicate library, and verifier semantics are not yet stable. Do not use for real CVE disclosure until v1.0 ships. See docs/SCOPE.md for the precise statement of what current bundles prove.

zkpox is a standalone framework for producing and verifying zero-knowledge proofs that you possess an exploit for a public program, without revealing the exploit.

It ships as two binaries:

Binary What it does
zkpox-prove Convert a target program + an exploit witness into a CBOR disclosure bundle.
zkpox-verify Validate a bundle end-to-end: STARK proof, target binding, envelope, anchor.

The proving backend is the SP1 zkVM. The disclosure pipeline layers:

  1. AES-256-GCM under a fresh key K over the witness bytes.
  2. age wrap of K to the vendor's public key (vendor-readable now).
  3. Drand tlock wrap of K to a future round (publicly readable after T).
  4. Sigstore Rekor anchor binding the bundle's hash to a time.

So the public proof is verifiable by anyone; the vendor learns the exploit immediately; the public learns the exploit at T (default 90 days, matching Project Zero's CVD norm).

Quickstart

# 1. Install the SP1 toolchain (https://docs.succinct.xyz/getting-started/install).
curl -L https://sp1up.succinct.xyz | bash && sp1up

# 2. Build the native binaries.
cargo build --release

# 3. Pre-build a backend for a specific target + predicate.
./target/release/zkpox-prove build-target \
    --target targets/03-libxml2-cve-2017-9047.c \
    --predicate memory-safety::oob-write

# 4. Produce a bundle (witness stays private; bundle is publishable).
./target/release/zkpox-prove prove \
    --target targets/03-libxml2-cve-2017-9047.c \
    --predicate memory-safety::oob-write \
    --witness tests/corpus/03-overflow1-crash.bin \
    --wrap groth16 \
    --output /tmp/bundle.cbor

# 5. Verify it. Strict is the default; every check must complete.
./target/release/zkpox-verify /tmp/bundle.cbor

# Optionally pin the Rekor log's public key to also verify the
# Signed Entry Timestamp (proves the log endorsed the entry):
./target/release/zkpox-verify /tmp/bundle.cbor --rekor-pubkey rekor.pub.pem

What this proves

For each (target program, predicate) pair, a zkpox-prove bundle establishes the statement:

"Under SP1's STARK verifier, I know an input w such that running the public target program on w causes the public predicate's vuln_flag to fire without triggering any inv_flag. The proof commits the hash of the target, the predicate ID and version, and the predicate's structured outputs (e.g. byte count and first offset of an out-of-bounds write). w itself is never disclosed by the proof."

What it does not prove: control-flow hijack, code execution, exploit reliability under ASLR/CET/canaries, or anything about the vulnerability class label — those are operator metadata. See docs/SCOPE.md for the full statement.

Architecture

flowchart TD
    target([target source])
    witness([witness])

    subgraph backend ["Backend layer"]
        direction LR
        l1["<b>Layer 1</b><br/>static-link C<br/><i>WORKING</i>"]
        l2["<b>Layer 2</b><br/>RV64IM emulator<br/><i>stub</i>"]
        l3["<b>Layer 3</b><br/>LLVM IR / MicroRAM<br/>CHEESECLOTH-style<br/><i>stub</i>"]
    end

    subgraph guest ["SP1 guest"]
        direction TB
        run["predicate runs against target"]
        commits["commits public values:<br/>target_hash, predicate,<br/>inv_flag, vuln_flag, outputs"]
        proof["STARK proof + public values"]
        run --> commits --> proof
    end

    subgraph bundle ["Bundle assembly"]
        direction TB
        env["envelope<br/>AES-GCM + age + Drand tlock"]
        anchor["Sigstore Rekor anchor"]
        cbor["CBOR encode"]
        env --> anchor --> cbor
    end

    final[/"bundle.cbor<br/><i>publishable</i>"/]

    target --> backend
    backend --> guest
    witness --> guest
    guest --> bundle
    bundle --> final

    classDef working fill:#2d6a4f,stroke:#52b788,color:#fff;
    classDef stub fill:#3a3a3a,stroke:#777,color:#bbb,stroke-dasharray:4 3;
    class l1 working;
    class l2,l3 stub;
Loading

Documentation

Document Purpose
docs/DESIGN.md Architecture + literature anchoring. CHEESECLOTH, Trail of Bits / SIEVE, SoK-on-SNARKs alignment.
docs/SCOPE.md What the proof asserts and what it does not. Read this first.
docs/PREDICATES.md The predicate library: catalogue, semantics, false-positive / false-negative profile, how to add a new one.
docs/BUNDLE-FORMAT.md CBOR schema reference.
docs/THREAT-MODEL.md Who trusts what; failure modes.
docs/DISCLOSURE-WORKFLOW.md Using zkpox for real CVD.
docs/ROADMAP.md Layer 2 / Layer 3 plans; predicate roadmap.

Acknowledgements

This work builds directly on:

  • CHEESECLOTH — Cuéllar, Harris, Parker, Pernsteiner, Tromer, USENIX Security 2023 / ACM TOPS 2025. The two-flag (inv_flag, vuln_flag) discipline and the memory-safety predicate catalogue come from CHEESECLOTH.
  • Trail of Bits + DARPA SIEVE — Bain et al., eprint 2022/1223 and the Trail of Bits blog series on ZK proof of exploit.
  • SP1 — Succinct's STARK-based zkVM, audited by Veridise, Cantina, Zellic, KALOS.
  • Drand — distributed randomness beacon used for the time-lock.
  • Sigstore Rekor — append-only transparency log for the timestamp anchor.
  • age and tlock — public-key and time-lock encryption tools.

Derived from prior work in gadievron/raptor#470.

License

Licensed under MIT.