Skip to content
KitploitKITPLOIT
ToolsExploitsBlog
Log in
Submit
ToolsExploitsBlog
Submit

Hacking, PenTest, and Cybersecurity Tools for Your Security Arsenal!

Kitploit is a directory of hacking, cybersecurity, and pentesting tools. Discover the latest project updates to find vulnerabilities, analyze systems, automate testing, and strengthen your security.

FeedsContactPrivacy© 2026 Kitploit

Tool Directory

Categories

View all categories
Loading categories
Tools/GitHubGitHub/anekazek/cve-2026-47627
Vulnerability AnalysisExploitationWeb Application ExploitationPenetration TestingLearning & EducationLabs & Practice
GitHubanekazek/cve-2026-47627

cve-2026-47627

Proof-of-concept exploit for CVE-2026-47627, a path traversal vulnerability in NVIDIA Triton Inference Server leading to arbitrary file write via Zip-Slip. Includes detailed analysis, lab environment, and one-click exploit scripts.

View Repository
3472 months agoNot yet reviewed
Share

CVE-2026-47627 — NVIDIA Triton Inference Server Path Traversal → DoS (Zip-Slip)

Complete research writeup & PoC in this repo (image nvcr.io/nvidia/tritonserver:26.05-py3 / server v2.69.0). Repo is set up to always be explicit so PoC works without restart via gRPC.


Table of Contents

  1. CVE Summary
  2. Lab Environment
  3. Vulnerability Anatomy
  4. Root Cause Analysis (Diff 26.05 → 26.06)
  5. Exploitation Paths: Which Ones Are Actually Affected?
  6. Main PoC: Zip-Slip via EXECUTION_ENV_PATH
  7. How to Run the PoC (One-Click)
  8. Traversal Evidence (Not a Fake 500)
  9. DoS Impact
  10. Mitigation
  11. Repo Structure (After Cleanup)
  12. References & Timeline

1. CVE Summary

FieldNilai
CVECVE-2026-47627
CNA[email protected]
Publish2026-08-18 (NVD Awaiting Analysis)
BulletinNV 5865 → https://github.com/NVIDIA/product-security/tree/main/2026/5865
CVSS 3.19.8 CRITICAL AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
CWECWE-22 Path Traversal
Affected0.0-26.05 (server v2.69.0 / core r26.05)
Fixed26.06 (server v2.70.0 / core r26.06)
ImpactDoS (deskripsi CNA) — riset ini membuktikan file write outside /models via Zip-Slip, yang bisa diperluas ke DoS/resource exhaustion
CreditMartin Brodeur

Honest note: The CNA description is only path traversal → DoS without details. This writeup reconstructs the bug location via diff r26.05..r26.06 and direct testing in the 26.05-py3 container.


2. Lab Environment

Image: nvcr.io/nvidia/tritonserver:26.05-py3 (TRITON_VERSION 2.69.0)

Compose is always explicit (already patched in this repo):

# docker-compose.yml:12
command: ["tritonserver", "--model-repository=/models", "--model-control-mode=explicit", "--log-verbose=0"]

run_triton.bat:10, run_triton.ps1:11, run_triton.sh:12 are also already explicit.

Triton Mode:

  • NONE (default upstream) → load once at start, POST /v2/repository/models/.../load → 503, must restart to load a new model.
  • POLL → scan repo every second, auto-load, still blocks explicit load.
  • EXPLICIT (this repo) → no scanning, but opens gRPC load_model / POST /load → PoC without restart via gRPC.

Port: | 8000 | HTTP REST | v2/health, v2/models, v2/repository | | 8001 | gRPC | RepositoryModelLoad, ModelInfer | | 8002 | Metrics | Prometheus |

Initial model:

model_repository/
├── echo_python/ (python backend, model.py)
│   └── 1/model.py
└── identity_onnx/ (onnxruntime, model.onnx ~1KB)

Quickstart:

docker compose up -d
docker compose logs -f
py scripts/client_test.py  # health + infer identity_onnx

3. Anatomy of the Vulnerability

User input (model_name / tar entry) ──► join(parent, child) ──► canonicalize ──► cek "child di dalam parent?" ──► buka file
         │                                    │                     │                         │                     │
         └──► "../tmp/pwn"               /models + "/../tmp/pwn" = /models/../tmp/pwn → /tmp/pwn    rfind vs find + '/' check    FileExists / extract
  • If join is used without canonicalize or with an incorrect rfind(parent,0)==0 check (partial match /models vs /models_evil), then ../ can escape.
  • DoS occurs if the traversed file is a FIFO, /dev/zero, /proc/self/mem, or a tar containing ../../ that overwrites system files → hang / OOM / crash.

Three research questions (from the initial report):

  1. Where does the user control the path? → model_name in the gRPC RepositoryModelLoad, EXECUTION_ENV_PATH tar entry, file: override, TRITON_BATCH_STRATEGY_PATH.
  2. How is the path used? → posixpath.join(repo, model_name) / temp_dir + "/" + file_name + IsChildPathEscapingParentPath.
  3. What is the impact? → DoS via resource exhaustion / hang. In this repo it is proven to cause a file write to /tmp/poc_marker.

4. Root Cause Analysis (Diff 26.05 → 26.06)

Repo core (r26.05..r26.06):

  1. src/filesystem/api.cc:407 IsChildPathEscapingParentPath — fix dcb315d + 72f3d9b:

    // VULN (r26.05):
    absolute_child.rfind(absolute_parent,0) != 0
    // → bug partial-match: "/models" prefix dari "/models_evil/file" dianggap inside
    
    // FIXED (r26.06):
    canonical_child.find(canonical_parent,0)==0 &&
    ((child.size() > parent.size() && child[parent.size()]=='/') || child.size()==parent.size())
    // → harus '/' setelah prefix, "/models_evil" sekarang correctly outside
    

    Used in src/backend_model.cc:196 (TRITON_BATCH_STRATEGY_PATH) and src/model_repository_manager/model_repository_manager.cc:162 (file:).

  2. src/model_repository_manager/model_repository_manager.cc:62 ValidateModelName — core#472/#481 (March 2026, already in 26.05 but tightened in 26.06):

    if (trimmed==".." || trimmed.find('/')!=npos) return INVALID_ARG "must not contain path traversal"
    

    This is what makes gRPC load_model('../tmp/pwn') → 400 INVALID_ARGUMENT in 26.05 (already blocked). But %2f (encoded /) slips through because it isn't a literal / → 500 literal poll.

  3. python_be.cc:292 EXECUTION_ENV_PATH — e520f8c7 test zipslip_test.py: The tar is extracted without checking for .. / absolute. The entry ../../poc_marker from /models/poc_exploit/malicious_env.tar.gz is extracted to /tmp/poc_marker (outside the model dir). Fix in 26.06: use ARCHIVE_EXTRACT_SECURE_NODOTDOT|NOABSOLUTEPATHS + check IsChildPathEscapingParentPath.

Repo server (r26.05..r26.06): only sagemaker_server.cc:1027 (RE2::FullMatch check) + http_server.cc:2440 (atoi→stoi), not the path for this CVE.

Conclusion: The model_name path via HTTP/gRPC is already blocked in 26.05 by ValidateModelName, but tar Zip-Slip is not — this is what the PoC in this repo exploits.


5. Exploitation Path: Which One Actually Hits?

VectorPayload26.05 ResultHit?
HTTP raw ../POST /v2/repository/models/../tmp/pwn/load404 (routing normalize)❌
HTTP encoded ..%2fPOST /v2/repository/models/..%2ftmp%2fpwn/load500 literal, poll /models/..%2ftmp%2fpwn doesn't exist❌ (fake 500, not traversal)
gRPC raw ../tmp/pwnload_model('../tmp/pwn')400 INVALID_ARGUMENT❌ (already blocked)
gRPC encoded ..%2fload_model('..%2ftmp%2fpwn')500 literal poll❌
file: override ../models_evilfile:../models_evil/pwn400/500 depending on partial-match⚠️ (depends on IsChildPathEscapingParentPath bug, but already blocked by ValidateModelName for model_name)
EXECUTION_ENV_PATH tar ../../poc_markermalicious_env.tar.gz entry ../../poc_marker200 + file at /tmp/poc_marker✅ VULN

The 500 vs 400 analogy in the old writeup is wrong: 500 = file doesn't exist (no door), 400 = rejected by validation (door locked). Neither gets in. Only 200 + a file at /tmp is proof.


6. Main PoC: Zip-Slip via EXECUTION_ENV_PATH

Idea: The Python backend parameter EXECUTION_ENV_PATH points to $$TRITON_MODEL_DIRECTORY/malicious_env.tar.gz. When load_model('poc_exploit') is called, the backend pb_env.cc:292 extracts the tar without sanitization. The entry ../../poc_marker escapes from .../poc_exploit/ to /tmp/poc_marker.

Download Tool

Most Popular

View all →

Discover the most used tools by our community.

  • 1

    OsintHunter

    letox4/osinthunter

    🎯 The ultimate OSINT toolkit: investigation graph, confidence scoring, alias families, map, timeline and reports. 49 modules, 400+ sites, no API keys required.

    930
  • 2

    ghostlock-app

    yukonga/ghostlock-app

    GhostLock One-Tap Execution App (CVE-2026-43499)

    554
  • 3

    redforge

    cybersecurity91/redforge

    Browser-side red team toolkit with 102 offline tools for recon, payload generation, credential attacks, crypto, encoding and engagement reporting.

    354
  • 4

    ARTEX

    baarai/artex

    LLM multi-agent autonomous penetration testing platform with planner/worker agents, asset and exploration graphs, MITM traffic recording, and vulnerability retesting.

    349
  • 5

    looking-glass

    anonghosty/looking-glass

    CSIRT platform that ingests Shadowserver reports into MongoDB and replays them as interactive maps, analysis, findings, advisories and executive dashboards.

    326
  • 6

    CyberStrikeAI

    ed1s0nz/cyberstrikeai

    The system of action for AI-native cybersecurity—where intent becomes governed execution, evidence becomes operational memory, and every operation improves the next.

    289

Explore all tools

Browse our collection of tools

View all tools →