CVE-2026-10546: Langflow SSRF via DNS rebinding

CISO Take

IBM Langflow's URL-loading component contains a Server-Side Request Forgery flaw rooted in a Time-of-Check/Time-of-Use race condition, allowing an attacker to bypass hostname validation via DNS rebinding and force the server to fetch attacker-chosen internal resources. The high confidentiality impact (C:H) means this can be used to reach cloud metadata endpoints, internal APIs, or other services on the Langflow host's network that were never meant to be internet-reachable, and no authentication is required to trigger the request. Exploitation requires high attack complexity — the attacker must win a DNS-rebinding race — which keeps the EPSS score low (0.00146) and explains why it isn't in CISA KEV and has no public exploit or Nuclei template yet; treat that as a temporary state rather than a reason to deprioritize, since SSRF-via-DNS-rebinding is a well-understood technique with mature public tooling. Any team running Langflow to build LLM pipelines that ingest URLs (web loaders, RAG document fetchers, tool integrations) should check IBM's advisory for a patched version, restrict outbound network access from the Langflow host with an egress allowlist, and disable or sandbox the URL data-source component until patched. Detection teams should watch for repeated/rapid DNS resolution changes for the same hostname immediately preceding outbound requests from the Langflow service.

Sources:NVDEPSSATLAS

What is the risk?

Medium overall risk (CVSS 6.5): the vulnerability is network-reachable and needs no privileges or user interaction, but the attack complexity is high because it depends on winning a TOCTOU race via DNS rebinding, which limits reliability and mass-exploitability. Impact is meaningful — high confidentiality loss from SSRF typically enables access to cloud metadata services (credential theft), internal admin panels, or other services segmented behind the Langflow host. Absence from CISA KEV, no public PoC, and low EPSS (bottom of exploited-in-the-wild range despite a 96th percentile ranking within EPSS's own distribution) indicate this is not yet under active attack, but SSRF/DNS-rebinding chains are a known, well-documented attack class with existing public tooling, so the barrier to a working exploit is more about effort than novel research.

How does the attack unfold?

Reconnaissance & setup

Attacker identifies a Langflow deployment with an exposed URL data-source component and registers a domain they control for DNS rebinding.

AML.T0006

Initial access via public-facing exploitation

Attacker submits the attacker-controlled URL into a Langflow flow, triggering the URL component's fetch logic.

AML.T0049

TOCTOU exploitation via DNS rebinding

Attacker rebinds DNS from a benign IP (passes validation) to an internal/restricted IP before the actual fetch occurs, exploiting the race condition.

Impact: internal data disclosure

Langflow's backend issues the request to the internal target on the attacker's behalf, and the response (e.g., metadata credentials or internal service data) is exposed back to the attacker through the flow output.

Reconnaissance & setup

Attacker identifies a Langflow deployment with an exposed URL data-source component and registers a domain they control for DNS rebinding.

AML.T0006

Initial access via public-facing exploitation

Attacker submits the attacker-controlled URL into a Langflow flow, triggering the URL component's fetch logic.

AML.T0049

TOCTOU exploitation via DNS rebinding

Attacker rebinds DNS from a benign IP (passes validation) to an internal/restricted IP before the actual fetch occurs, exploiting the race condition.

Impact: internal data disclosure

Langflow's backend issues the request to the internal target on the attacker's behalf, and the response (e.g., metadata credentials or internal service data) is exposed back to the attacker through the flow output.

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
Langflow	pip	—	No patch
156.2KPushed today18% patched~66d to patchFull package profile →

Do you use Langflow? You're affected.

How severe is it?

CVSS 3.1

6.5 / 10

EPSS

0.2%

chance of exploitation in 30 days

Higher than 16% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Moderate

What is the attack surface?

AVNetwork

ACHigh

PRNone

UINone

SUnchanged

CHigh

ILow

ANone

What should I do?

1 step

Upgrade to the patched Langflow release referenced in IBM's advisory as soon as it's available for your version line; until then, disable or restrict the URL/data_source component for non-trusted users. Apply network-level egress controls (firewall/proxy allowlists) so the Langflow host cannot reach internal-only IP ranges, link-local addresses (169.254.0.0/16, incl. cloud metadata endpoints), or other sensitive internal services. Where feasible, front outbound HTTP fetches with a validating forward proxy that re-resolves and re-checks the destination IP at request time (mitigating the TOCTOU gap) rather than trusting a single validate-then-connect check. For detection, monitor Langflow's outbound traffic for requests to internal/RFC1918 or link-local destinations, and flag hostnames whose DNS resolution changes shortly before a fetch (classic DNS-rebinding signature).

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 LeakageCode ExecutionFrameworkAgentAML.T0010.005- AI Agent Tool AML.T0049- Exploit Public-Facing Application

Which compliance frameworks are affected?

This CVE is relevant to:

ISO 42001

A.6.2.4-AI system operation and monitoring

NIST AI RMF

MANAGE-2.3-Third-party AI system risks are managed

OWASP LLM Top 10

LLM08-Excessive Agency / Insecure Plugin Design

Frequently Asked Questions

What is CVE-2026-10546?

IBM Langflow's URL-loading component contains a Server-Side Request Forgery flaw rooted in a Time-of-Check/Time-of-Use race condition, allowing an attacker to bypass hostname validation via DNS rebinding and force the server to fetch attacker-chosen internal resources. The high confidentiality impact (C:H) means this can be used to reach cloud metadata endpoints, internal APIs, or other services on the Langflow host's network that were never meant to be internet-reachable, and no authentication is required to trigger the request. Exploitation requires high attack complexity — the attacker must win a DNS-rebinding race — which keeps the EPSS score low (0.00146) and explains why it isn't in CISA KEV and has no public exploit or Nuclei template yet; treat that as a temporary state rather than a reason to deprioritize, since SSRF-via-DNS-rebinding is a well-understood technique with mature public tooling. Any team running Langflow to build LLM pipelines that ingest URLs (web loaders, RAG document fetchers, tool integrations) should check IBM's advisory for a patched version, restrict outbound network access from the Langflow host with an egress allowlist, and disable or sandbox the URL data-source component until patched. Detection teams should watch for repeated/rapid DNS resolution changes for the same hostname immediately preceding outbound requests from the Langflow service.

Is CVE-2026-10546 actively exploited?

No confirmed active exploitation of CVE-2026-10546 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-10546?

Upgrade to the patched Langflow release referenced in IBM's advisory as soon as it's available for your version line; until then, disable or restrict the URL/data_source component for non-trusted users. Apply network-level egress controls (firewall/proxy allowlists) so the Langflow host cannot reach internal-only IP ranges, link-local addresses (169.254.0.0/16, incl. cloud metadata endpoints), or other sensitive internal services. Where feasible, front outbound HTTP fetches with a validating forward proxy that re-resolves and re-checks the destination IP at request time (mitigating the TOCTOU gap) rather than trusting a single validate-then-connect check. For detection, monitor Langflow's outbound traffic for requests to internal/RFC1918 or link-local destinations, and flag hostnames whose DNS resolution changes shortly before a fetch (classic DNS-rebinding signature).

What systems are affected by CVE-2026-10546?

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

What is the CVSS score for CVE-2026-10546?

CVE-2026-10546 has a CVSS v3.1 base score of 6.5 (MEDIUM). The EPSS exploitation probability is 0.25%.

What is the AI security impact?

Affected AI Architectures

agent frameworksRAG pipelinesLLM orchestration/flow builders

MITRE ATLAS Techniques

AML.T0010.005AI Agent Tool

AML.T0049Exploit Public-Facing Application

Compliance Controls Affected

ISO 42001: A.6.2.4

NIST AI RMF: MANAGE-2.3

OWASP LLM Top 10: LLM08

What are the technical details?

Original Advisory

IBM Langflow OSS 1.0.0 through 1.9.3 contains a Server-Side Request Forgery (SSRF) vulnerability in the URL component ( src/lfx/src/lfx/components/data_source/url.py ) due to a Time-of-Check/Time-of-Use (TOCTOU) race condition that can be exploited via DNS rebinding.

Exploitation Scenario

An attacker with access to a Langflow flow builder (or the ability to submit a URL into a flow that includes the URL data-source component) registers a domain whose DNS record they control. They first resolve the domain to a benign, allow-listed IP so Langflow's initial validation check passes, then rapidly rebind the DNS record to an internal IP (e.g., the cloud metadata service or an internal admin API) before Langflow actually issues the fetch — exploiting the gap between check and use. Langflow's backend then makes the HTTP request on the attacker's behalf to the internal target, returning the response (e.g., cloud credentials or internal service data) back into the flow's output, which the attacker can read via the UI or API, achieving unauthorized read access to internal-only resources.

Weaknesses (CWE)

CWE-918Server-Side Request ForgeryPrimary CWE-918Server-Side Request Forgery

CWE-918 — Server-Side Request Forgery: The web server receives a URL or similar request from an upstream component and retrieves the contents of this URL, but it does not sufficiently ensure that the request is being sent to the expected destination.

[Implementation, Operation] If there is a known list of applications, IP, hostname, and/or domains to which the product is allowed to communicate, then use allowlists to ensure that outgoing requests are restricted to these known entities. The allowlists could be in the application code or possibly enforced by firewalls or network segmentation [REF-1511].
[Implementation, Operation] If the product is expected to be able to connect to arbitrary IP addresses or domains, consider using a denylist that includes private network addresses, internal domain names, etc. The denylist could be in the application code or possibly enforced by firewalls or network segmentation [REF-1511].

Source: MITRE CWE corpus.