CVE-2026-17621: Langflow path traversal

CISO Take

IBM Langflow OSS versions 1.0.0 through 1.10.2 contain a classic directory traversal flaw (CWE-22): a low-privileged, network-reachable attacker can send a URL with '../' sequences to read arbitrary files on the host, including configuration files, environment variables, and credentials. Langflow is a visual builder for LLM and agent pipelines, so any exposed instance is a likely repository of API keys for LLM providers, vector database credentials, and internal flow definitions — a single unauthenticated-adjacent read can cascade into full compromise of the AI stack behind it. There is no CISA KEV listing, no published EPSS score, and no known public exploit or scanner template, so this looks opportunistic rather than actively targeted today, but the CVSS 5.4 (AV:N/AC:L/PR:L/UI:N) profile means it requires only low privileges and no user interaction to trigger. Patch to the fixed IBM Langflow release referenced in the vendor advisory as soon as it's available, and in the meantime restrict network exposure of Langflow instances, put them behind an authenticating reverse proxy, and add WAF/log monitoring for '../' patterns in requests. Treat any secrets accessible to the Langflow process (API keys, DB creds) as potentially exposed and rotate them if the instance has been internet-facing.

Sources:NVDibm.comATLAS

What is the risk?

Medium severity per CVSS (5.4) reflecting a confidentiality-only impact (C:L/I:N/A:N) with low attack complexity and no user interaction, but requiring low privileges (an authenticated account on the Langflow instance). No EPSS score, no CISA KEV entry, and no public exploit code or Nuclei template have been observed, indicating this is not yet under active mass exploitation. However, the low bar for exploitation (a crafted URL) combined with network exposure of many self-hosted Langflow deployments (often run with broad file-system access for flow storage and package management) means real-world exploitability is higher than the raw CVSS score suggests once an attacker has any foothold with a low-privilege account.

How does the attack unfold?

Initial Access

Attacker obtains a low-privileged account or session on a network-exposed Langflow instance.

AML.T0012

Exploitation

Attacker sends a crafted URL request with '../' sequences to a Langflow endpoint to traverse outside the intended directory.

AML.T0049

Data Access

Attacker reads arbitrary host files such as '.env' configuration or flow definitions containing LLM API keys and database credentials.

AML.T0037

Impact / Escalation

Attacker uses the harvested credentials to access connected LLM APIs or vector databases, expanding the compromise beyond the initial file read.

AML.T0040

Initial Access

Attacker obtains a low-privileged account or session on a network-exposed Langflow instance.

AML.T0012

Exploitation

Attacker sends a crafted URL request with '../' sequences to a Langflow endpoint to traverse outside the intended directory.

AML.T0049

Data Access

Attacker reads arbitrary host files such as '.env' configuration or flow definitions containing LLM API keys and database credentials.

AML.T0037

Impact / Escalation

Attacker uses the harvested credentials to access connected LLM APIs or vector databases, expanding the compromise beyond the initial file read.

AML.T0040

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Langflow	pip	—	No patch
156.2KPushed yesterday18% patched~66d to patchFull package profile →

Do you use Langflow? You're affected.

How severe is it?

CVSS 3.1

5.4 / 10

EPSS

0.4%

chance of exploitation in 30 days

Higher than 34% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Trivial

What is the attack surface?

AVNetwork

ACLow

PRLow

UINone

SUnchanged

CLow

ILow

ANone

What should I do?

1 step

1) Upgrade Langflow to the version that remediates this issue per the IBM support advisory (linked reference) as soon as it is published/available — verify the fixed version number directly from IBM's advisory since it is not stated in the CVE description. 2) Until patched, do not expose Langflow directly to the internet; place it behind a reverse proxy or VPN with authentication, and restrict access to trusted IP ranges. 3) Add WAF/IDS rules to detect and block URL-encoded and raw '../' traversal sequences in requests to the Langflow service. 4) Audit and rotate any credentials (LLM API keys, DB connection strings) that were reachable from the Langflow host's filesystem, especially for internet-facing deployments. 5) Review Langflow access logs for anomalous requests containing dot-dot-slash patterns or attempts to reach '/etc/passwd', '.env', or similar sensitive paths. 6) Apply least-privilege file-system permissions to the Langflow service account so even a successful traversal exposes minimal sensitive content.

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?

Data LeakageFrameworkAgentAML.T0007- Discover AI Artifacts AML.T0037- Data from Local System AML.T0049- Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Article 15-Accuracy, robustness and cybersecurity

ISO 42001

A.6.2.6-AI system security

NIST AI RMF

MEASURE 2.7-AI system security and resilience evaluated

OWASP LLM Top 10

LLM02:2025-Sensitive Information Disclosure

Frequently Asked Questions

What is CVE-2026-17621?

IBM Langflow OSS versions 1.0.0 through 1.10.2 contain a classic directory traversal flaw (CWE-22): a low-privileged, network-reachable attacker can send a URL with '../' sequences to read arbitrary files on the host, including configuration files, environment variables, and credentials. Langflow is a visual builder for LLM and agent pipelines, so any exposed instance is a likely repository of API keys for LLM providers, vector database credentials, and internal flow definitions — a single unauthenticated-adjacent read can cascade into full compromise of the AI stack behind it. There is no CISA KEV listing, no published EPSS score, and no known public exploit or scanner template, so this looks opportunistic rather than actively targeted today, but the CVSS 5.4 (AV:N/AC:L/PR:L/UI:N) profile means it requires only low privileges and no user interaction to trigger. Patch to the fixed IBM Langflow release referenced in the vendor advisory as soon as it's available, and in the meantime restrict network exposure of Langflow instances, put them behind an authenticating reverse proxy, and add WAF/log monitoring for '../' patterns in requests. Treat any secrets accessible to the Langflow process (API keys, DB creds) as potentially exposed and rotate them if the instance has been internet-facing.

Is CVE-2026-17621 actively exploited?

No confirmed active exploitation of CVE-2026-17621 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-17621?

1) Upgrade Langflow to the version that remediates this issue per the IBM support advisory (linked reference) as soon as it is published/available — verify the fixed version number directly from IBM's advisory since it is not stated in the CVE description. 2) Until patched, do not expose Langflow directly to the internet; place it behind a reverse proxy or VPN with authentication, and restrict access to trusted IP ranges. 3) Add WAF/IDS rules to detect and block URL-encoded and raw '../' traversal sequences in requests to the Langflow service. 4) Audit and rotate any credentials (LLM API keys, DB connection strings) that were reachable from the Langflow host's filesystem, especially for internet-facing deployments. 5) Review Langflow access logs for anomalous requests containing dot-dot-slash patterns or attempts to reach '/etc/passwd', '.env', or similar sensitive paths. 6) Apply least-privilege file-system permissions to the Langflow service account so even a successful traversal exposes minimal sensitive content.

What systems are affected by CVE-2026-17621?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, RAG pipelines, LLM orchestration/workflow builders.

What is the CVSS score for CVE-2026-17621?

CVE-2026-17621 has a CVSS v3.1 base score of 5.4 (MEDIUM). The EPSS exploitation probability is 0.40%.

What is the AI security impact?

Affected AI Architectures

agent frameworksRAG pipelinesLLM orchestration/workflow builders

MITRE ATLAS Techniques

AML.T0007Discover AI Artifacts

AML.T0037Data from Local System

AML.T0049Exploit Public-Facing Application

Compliance Controls Affected

EU AI Act: Article 15

ISO 42001: A.6.2.6

NIST AI RMF: MEASURE 2.7

OWASP LLM Top 10: LLM02:2025

What are the technical details?

Original Advisory

IBM Langflow OSS 1.0.0 through 1.10.2 could allow a remote attacker to traverse directories on the system. An attacker could send a specially crafted URL request containing "dot dot " sequences ( /.. /) to view arbitrary files on the system.

Exploitation Scenario

An attacker who has obtained a low-privileged account on a self-hosted Langflow instance (e.g., via a shared trial, weak signup controls, or a phished credential) sends a specially crafted HTTP request containing '../' sequences to a file-access endpoint. The traversal lets them read files outside the intended directory, such as the Langflow '.env' file containing the OpenAI/Anthropic API key and vector database credentials used by the organization's RAG pipelines. The attacker then uses the harvested LLM API key to run unauthorized inference at the victim's expense, or uses the vector DB credentials to exfiltrate proprietary embeddings and documents indexed for retrieval-augmented generation.

Weaknesses (CWE)

CWE-22Improper Limitation of a Pathname to a Restricted DirectoryPrimary CWE-22Improper Limitation of a Pathname to a Restricted Directory

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.