CVE-2026-103254: n8n: path traversal enables cross-project approval forgery

MEDIUM
Published October 1, 2026
CISO Take

A path traversal flaw in n8n's signed resume URL generation for Send-and-Wait approvals lets a user with workflow-creation rights mint valid approval URLs for human-in-the-loop gates in projects they cannot access. The practical risk is approval forgery: an attacker could self-approve actions that a different team intended a human to authorize, such as agent tool calls, data exports or financial steps. Exploitation is rated high complexity and needs an authenticated insider or compromised low-privilege account. EPSS is 0.0028 (top 81% percentile), there is no public exploit or Nuclei template, and the CVE is not in CISA KEV, so this is a scheduled-patching item rather than an emergency. Upgrade to n8n 1.123.80, 2.39.6 or 2.40.1 (or later), and review which workflows use approval gates to protect sensitive agent actions.

Sources:NVDEPSSGitHub AdvisoryVulnCheckATLAS

What is the risk?

Medium (CVSS 6.3). The attack is network-based but needs high complexity and low privileges (workflow creation). Scope is changed because it crosses project boundaries, and integrity impact is high with no confidentiality or availability impact. Exploitation likelihood is low: EPSS 0.0028, no public PoC, no scanner template, not in KEV. Risk rises in multi-tenant or multi-project n8n instances where many users hold workflow-creation rights and approval gates guard AI agent actions with real-world side effects.

How does the attack unfold?

Initial Access
Attacker uses a low-privilege account with workflow-creation rights in one n8n project.
AML.T0012
Exploitation
Attacker supplies a node ID with traversal sequences so the signed resume URL resolves to a gate in another project.
AML.T0049
Approval Forgery
Attacker calls the minted valid URL to approve a pending execution in the victim project.
AML.T0053
Impact
The gated AI agent or workflow action runs without legitimate human authorization, compromising integrity.

What systems are affected?

PackageEcosystemVulnerable RangePatched
n8nnpm—No patch
210.5KOpenSSF 7.116 dependentsPushed 3d ago42% patched~19d to patchFull package profile →

Do you use n8n? You're affected.

How severe is it?

CVSS 3.1
6.3 / 10
EPSS
0.3%
chance of exploitation in 30 days
Higher than 21% of all CVEs
Exploitation Status
No known exploitation
Sophistication
Moderate

What is the attack surface?

AVACPRUISCIA
AVNetwork
ACHigh
PRLow
UINone
SChanged
CNone
IHigh
ANone

What should I do?

1 step
  1. 1) Upgrade n8n to 1.123.80, 2.39.6, 2.40.1 or later depending on your release line. 2) Until patched, restrict workflow-creation permissions to trusted users and review who holds them across projects. 3) Audit workflows using Send-and-Wait and look for resume URLs or node IDs containing traversal sequences such as ../ or encoded variants. 4) Review execution logs for approvals resolved from unexpected users, IPs or projects, and for workflows created or edited with unusual node IDs. 5) For high-impact agent actions, add an independent control outside n8n (secondary approval, rate limits, least-privilege credentials). 6) Rotate credentials used by approval-gated workflows if forged approvals are suspected. Advisory: GHSA-597w-c3jh-g8fg.

What does CISA's SSVC say?

DecisionTrack
Exploitationnone
AutomatableNo
Technical Impactpartial

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act
Art. 14-Human oversight
ISO 42001
A.9.2-Processes for responsible use of AI systems
NIST AI RMF
GOVERN 1.4-Risk management process and controls
OWASP LLM Top 10
LLM06-Excessive Agency

Frequently Asked Questions

What is CVE-2026-103254?

A path traversal flaw in n8n's signed resume URL generation for Send-and-Wait approvals lets a user with workflow-creation rights mint valid approval URLs for human-in-the-loop gates in projects they cannot access. The practical risk is approval forgery: an attacker could self-approve actions that a different team intended a human to authorize, such as agent tool calls, data exports or financial steps. Exploitation is rated high complexity and needs an authenticated insider or compromised low-privilege account. EPSS is 0.0028 (top 81% percentile), there is no public exploit or Nuclei template, and the CVE is not in CISA KEV, so this is a scheduled-patching item rather than an emergency. Upgrade to n8n 1.123.80, 2.39.6 or 2.40.1 (or later), and review which workflows use approval gates to protect sensitive agent actions.

Is CVE-2026-103254 actively exploited?

No confirmed active exploitation of CVE-2026-103254 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-103254?

1) Upgrade n8n to 1.123.80, 2.39.6, 2.40.1 or later depending on your release line. 2) Until patched, restrict workflow-creation permissions to trusted users and review who holds them across projects. 3) Audit workflows using Send-and-Wait and look for resume URLs or node IDs containing traversal sequences such as ../ or encoded variants. 4) Review execution logs for approvals resolved from unexpected users, IPs or projects, and for workflows created or edited with unusual node IDs. 5) For high-impact agent actions, add an independent control outside n8n (secondary approval, rate limits, least-privilege credentials). 6) Rotate credentials used by approval-gated workflows if forged approvals are suspected. Advisory: GHSA-597w-c3jh-g8fg.

What systems are affected by CVE-2026-103254?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, workflow automation, human-in-the-loop approval gates, multi-project orchestration platforms.

What is the CVSS score for CVE-2026-103254?

CVE-2026-103254 has a CVSS v3.1 base score of 6.3 (MEDIUM). The EPSS exploitation probability is 0.28%.

What is the AI security impact?

Affected AI Architectures

agent frameworksworkflow automationhuman-in-the-loop approval gatesmulti-project orchestration platforms

MITRE ATLAS Techniques

AML.T0012Valid Accounts
AML.T0053AI Agent Tool Invocation

Compliance Controls Affected

EU AI Act: Art. 14
ISO 42001: A.9.2
NIST AI RMF: GOVERN 1.4
OWASP LLM Top 10: LLM06

What are the technical details?

Original Advisory

n8n versions before 1.123.80, from 2.0.0 before 2.39.6, and from 2.40.0 before 2.40.1 contain a path traversal vulnerability in signed resume URL generation for Send-and-Wait approvals. Attackers with workflow creation permissions can mint valid approval URLs for gates in projects they cannot access by exploiting unresolved traversal sequences in caller-controlled node IDs, enabling cross-project approval forgery.

Exploitation Scenario

An attacker with a low-privilege n8n account in Project A can create workflows. They craft a workflow whose Send-and-Wait node has a caller-controlled node ID containing traversal sequences that resolve to the node ID of an approval gate in Project B, such as a finance or production-ops agent that pauses for human review before executing a tool call. The platform signs a resume URL for that target path, and the attacker calls it to approve the pending execution without the legitimate reviewer's involvement. The AI agent then proceeds with the gated action, for example a data export, a payment or an infrastructure change, which the attacker was never authorized to approve.

Weaknesses (CWE)

CWE-22 — Improper Limitation of a Pathname to a Restricted Directory: The product uses external input to construct a pathname that is intended to identify a file or directory that is located underneath a restricted parent directory, but the product does not properly neutralize special elements within the pathname that can cause the pathname to resolve to a location that is outside of the restricted directory.

  • [Implementation] Assume all input is malicious. Use an "accept known good" input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does. When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, "boat" may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected to contain colors such as "red" or "blue." Do not rely exclusively on looking for malicious or malformed inputs. This is likely to miss at least one undesirable input, especially if the code's environment changes. This can give attackers enough room to bypass the intended validation. However, denylis
  • [Architecture and Design] For any security checks that are performed on the client side, ensure that these checks are duplicated on the server side, in order to avoid CWE-602. Attackers can bypass the client-side checks by modifying values after the checks have been performed, or by changing the client to remove the client-side checks entirely. Then, these modified values would be submitted to the server.

Source: MITRE CWE corpus.

CVSS Vector

CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:C/C:N/I:H/A:N

Timeline

Published
October 1, 2026
Last Modified
October 6, 2026
First Seen
October 1, 2026

Related Vulnerabilities