Skip to content
KitploitKITPLOIT
StrumentiExploitsBlog
Log in
Invia
StrumentiExploitsBlog
Invia

Strumenti di Hacking, PenTest e Cybersecurity per il tuo Arsenale di Sicurezza!

Kitploit è una directory di strumenti di hacking, cybersecurity e pentesting. Scopri gli ultimi aggiornamenti dei progetti per trovare vulnerabilità, analizzare sistemi, automatizzare i test e rafforzare la tua sicurezza.

FeedContattoPrivacy© 2026 Kitploit

Directory degli strumenti

Categorie

Vedi tutte le categorie
Loading categories
intbug-lab — Reproducible loopback lab + invariant battery for HTTP/2 flow-control integer bugs (nginx CVE-2026-123007 class). Lab-only, coordinated-disclosure. | Kitploit
Strumenti/GitHubGitHub/c0re-i5/intbug-lab
Defensive ToolsWeb Vulnerability ScannersContainer SecurityVulnerability AnalysisFuzzingNetwork SecurityPapers & ResearchLearning & EducationLabs & Practice
GitHubc0re-i5/intbug-lab

intbug-lab

Reproducible loopback lab + invariant battery for HTTP/2 flow-control integer bugs (nginx CVE-2026-123007 class). Lab-only, coordinated-disclosure.

22h 37m faNon ancora revisionato
Condividi
Vedi Repository
Contenuto non disponibile nella lingua richiesta. Visualizzazione della versione inglese.

HTTP/2 Flow-Control Integer-Bug Lab

Reproducible loopback lab + invariant battery for the HTTP/2 flow-control / counter / length integer-bug class (CWE-191), validated against nginx CVE-2026-123007.

Validated result: the cleanup_overcredit probe FAILs on nginx 1.31.8 (known-vulnerable) and PASSes on nginx 1.31.9 (patched) and on Apache, Caddy, Envoy, Go net/http, and nghttpd — matching arkrwn's public --safe PoC on the same listeners. Eight real HTTP/2 stacks; seven RFC 9113 probes; one-command run.

Longer write-up: WRITEUP.md.

  • Loopback only (127.0.0.1). Drivers refuse non-loopback hosts.
  • Lab-only, coordinated-disclosure. No internet scanning. No RCE. No ~119 GiB / 2^64 walk. Not tied to Jig or any other project.

Quick start

git clone https://github.com/c0re-i5/intbug-lab.git
cd intbug-lab
python3 -m venv .venv && .venv/bin/pip install -r ours/requirements.txt
./scripts/run-all.sh          # or: make battery

That brings up the pinned targets, runs offline unit tests, and writes the matrix under evidence/. Tear down with ./scripts/lab-down-targets.sh (or make down).

Optional: third-party ground truth (not required for the battery)

# arkrwn --safe PoC (for scripts/proof.sh cross-check)
git clone https://github.com/arkrwn/NGINX2Root-CVE-2026-123007.git \
  detectors/arkrwn-nginx2root
(cd detectors/arkrwn-nginx2root && cargo build --release)

# Rapid7 reproduction notes (reference only)
git clone https://github.com/sfewer-r7/CVE-2026-123007.git \
  detectors/sfewer-r7-cve

# h2spec for the optional zero-WU RFC cross-check (make zero-wu)
# see bin/README.md for the download command

Details: detectors/README.md, bin/README.md.

One-command entrypoint

make battery          # = ./scripts/run-all.sh
# or, skip the ~45s heavy probe:
make battery-light    # = ./scripts/run-all.sh --skip-heavy

./scripts/run-all.sh is the canonical one-command (also exposed as make battery when GNU make is installed). It is idempotent:

  1. ensure dockerd (./scripts/box-dockerd.sh)
  2. bring all pinned targets up (./scripts/lab-up-targets.sh)
  3. run offline unit tests
  4. run the 7-probe battery across every registered target and write the matrix

Exit codes: 0 after a successful harness run (FAIL/ANOMALY in the matrix are observations, not harness failures). Non-zero only if docker is unreachable, unit tests fail, or the battery crashes.

Tear down: make down (or ./scripts/lab-down-targets.sh).

Targets (pinned)

idRoleListenImage digest
nginx-1.31.8known-vulnerable (CVE-2026-123007)127.0.0.1:19443nginx@sha256:5ab5b0b01097420873a3f3330f44248e34a3c2a3e8d5ee30658aeae8cb7435da
nginx-http2-offnegative control (same image, http2 off)127.0.0.1:19444same as above
httpd-2.4.71cross-impl (Apache + mod_http2)127.0.0.1:19445httpd@sha256:32e9403360ab2f1047523e72f256527fb9aee68b342abf6ad03f0d629157f3d9
nginx-1.31.9patched clean control127.0.0.1:19446nginx@sha256:08f354a3b44c422021a5c0f5efb3e27f8f90236b5447d3bac4053f5f50ab6760
caddy-2.10cross-impl (Caddy)127.0.0.1:19447caddy@sha256:c3d7ee5d2b11f9dc54f947f68a734c84e9c9666c92c88a7f30b9cba5da182adb
envoy-1.32cross-impl (Envoy)127.0.0.1:19448envoyproxy/envoy@sha256:6bf0d37dd5e8eac4b37effe89a1aec4760f7f2ce860d1fd6af22147eb87a5058
go-nethttpcross-impl (Go net/http stdlib)127.0.0.1:19449built from golang@sha256:1a6d4452c65dea36aac2e2d606b01b4a029ec90cc1ae53890540ce6173ea77ac
nghttpdreference (nghttp2)127.0.0.1:19450built from debian@sha256:734a72b3d6142a71993b4bc580ca01d8ca687f77a2c422d96a961d3376b179bc (+ apt nghttp2-server, not itself digest-pinned)

All publishes are 127.0.0.1 only. Bodies are 896-byte /index.html (Range-capable).

Reproducibility limits (honest)

  • Official images (nginx / httpd / caddy / envoy / golang / debian) are digest-pinned.
  • go-nethttp is rebuilt from the pinned golang digest with stdlib only (no module fetch).
  • nghttpd installs nghttp2-server via apt on a pinned debian base — the apt package version floats with the distro snapshot. Treat it as a reference baseline, not a bit-for-bit pin. Documented here so nobody pretends otherwise.

Invariant battery (bug class, not one CVE)

Implemented in ours/battery.py. Each probe cites RFC 9113 (HTTP/2).

Probe idWhat it checksRFC
h2_alpnALPN h2 present/absent as expected§3.3 / RFC 7301
wu_zero_connWINDOW_UPDATE increment 0 on connection → PROTOCOL_ERROR§6.9.1
wu_zero_streamWINDOW_UPDATE increment 0 on stream → PROTOCOL_ERROR§6.9.1
wu_conn_overflowconnection WU that would exceed MAX=2^31-1 → FLOW_CONTROL_ERROR§6.9.1
wu_stream_overflowstream WU that would exceed MAX → FLOW_CONTROL_ERROR§6.9.1
settings_iw_overflowSETTINGS_INITIAL_WINDOW_SIZE delta pushing a stream past MAX§6.9.2
cleanup_overcreditqueued-DATA cleanup must not drive connection window above MAX; post-refund WU guard (CVE-2026-123007 accounting class)§6.9 / CWE-191

Statuses: PASS (invariant held), FAIL (invariant broken — bug-class signal), ANOMALY (needs manual review, not a confirmed CVE), INCONCLUSIVE (probe could not complete — not a PASS), SKIP, ERROR.

Heavy probe cleanup_overcredit ≈ 14 MiB / ~45s per target. Skip with ./scripts/run-battery.sh --skip-heavy or make battery-light.

Results matrix (this box, 2026-11-19 America/Toronto)

From evidence/BATTERY-MATRIX-latest.txt (graceful cleanup + open-stream wu_zero fix):

probe                | nginx-1.31.8    | nginx-http2-off | httpd-2.4.71    | nginx-1.31.9    | caddy-2.10      | envoy-1.32      | go-nethttp      | nghttpd
h2_alpn              | PASS            | PASS            | PASS            | PASS            | PASS            | PASS            | PASS            | PASS
wu_zero_conn         | PASS            | -               | PASS            | PASS            | PASS            | ANOMALY         | PASS            | PASS
wu_zero_stream       | PASS            | -               | PASS            | PASS            | ANOMALY         | ANOMALY         | ANOMALY         | PASS
wu_conn_overflow     | PASS            | -               | PASS            | PASS            | PASS            | ANOMALY         | PASS            | PASS
wu_stream_overflow   | FAIL            | -               | ANOMALY         | FAIL            | ANOMALY         | ANOMALY         | ANOMALY         | ANOMALY
settings_iw_overflow | FAIL            | -               | ANOMALY         | FAIL            | PASS            | ANOMALY         | PASS            | ANOMALY
cleanup_overcredit   | FAIL            | -               | PASS            | PASS            | PASS            | PASS            | PASS            | PASS

cleanup_overcredit now completes on all 8 h2 targets (no Broken-pipe soft-PASS). caddy/envoy/go/nghttpd use key_updates=0 / carrier=200000 (full KeyUpdate pressure hits Broken pipe on those stacks). See evidence/MANUAL-REVIEW-NOTES.txt.

wu_zero_stream prior FAILs on Apache/Envoy/nghttpd were a probe artifact (WU after END_STREAM). Open-stream (h2spec-aligned) probe + byte capture + h2spec http2/6.9/1+2 cross-check transcripts are under evidence/zero-wu/ (optional binary; see bin/README.md).

How to read the matrix (honesty rules)

Scarica lo strumento

Più Popolari

Vedi tutti →

Scopri gli strumenti più utilizzati dalla nostra community.

  • 1

    OsintHunter

    letox4/osinthunter

    🎯 Il toolkit OSINT definitivo: grafo di indagine, punteggio di affidabilità, famiglie di alias, mappa, timeline e report. 49 moduli, oltre 400 siti, nessuna chiave API richiesta.

    938
  • 2

    ghostlock-app

    yukonga/ghostlock-app

    # App di esecuzione con un tocco GhostLock (CVE-2026-43499)

    556
  • 3

    redforge

    cybersecurity91/redforge

    Toolkit red team lato browser con 102 strumenti offline per ricognizione, generazione di payload, attacchi alle credenziali, crittografia, codifica e reportistica dell'engagement.

    352
  • 4

    ARTEX

    baarai/artex

    Piattaforma di penetration testing autonoma multi-agente basata su LLM con agenti planner/worker, grafi di asset e di esplorazione, registrazione del traffico MITM e retesting delle vulnerabilità.

    343
  • 5

    looking-glass

    anonghosty/looking-glass

    Piattaforma CSIRT che importa i report di Shadowserver in MongoDB e li riproduce come mappe interattive, analisi, risultati, avvisi e dashboard executive.

    336
  • 6

    CyberStrikeAI

    ed1s0nz/cyberstrikeai

    Il sistema d'azione per la cybersecurity nativa AI—dove l'intento diventa esecuzione governata, le evidenze diventano memoria operativa, e ogni operazione migliora la successiva.

    291

Esplora tutti gli strumenti

Sfoglia la nostra collezione di strumenti

Vedi tutti gli strumenti →