CVE-2026-18640: path traversal corrupts ACL/hunt

CISO Take

A flaw in Velociraptor's NewNotebook API fails to properly sanitize file path parameters, letting any authenticated user who already holds the NOTEBOOK_EDIT permission write files outside the organization's designated data store directory. The forced ".json.db" extension limits the blast radius to metadata-type files, but that still includes access control lists and hunt records, so a malicious or compromised low-privilege user can corrupt the very files that govern who can see and do what inside the platform. This is not a remote-code-execution bug and there is no public exploit, no Nuclei template, and it is not in CISA KEV — CISA's own SSVC decision rates it TRACK, its lowest urgency tier, and EPSS puts real-world exploitation probability at just 0.3%. Because Velociraptor is frequently used as forensic and incident-response tooling, integrity damage to ACL or hunt metadata could quietly degrade an organization's ability to trust or reconstruct an active investigation, which raises the stakes above what the CVSS 7.1 score alone conveys. Patch to the fixed Velociraptor release per the vendor advisory, and in the meantime restrict NOTEBOOK_EDIT to trusted operators and monitor for unexpected writes to ACL/hunt metadata files as a detection signal.

Sources:NVDCISA KEVEPSSdocs.velociraptor.app

What is the risk?

CVSS 7.1 (High) reflects a network-reachable, low-complexity flaw with high integrity impact and low availability impact, but Privileges Required is Low — exploitation requires an already-authenticated account holding the specific NOTEBOOK_EDIT permission, which meaningfully narrows the attacker population versus an unauthenticated bug. No public exploit code, no Nuclei scanning template, absence from CISA KEV, and CISA's own TRACK SSVC decision (the lowest of its four tiers) all point to low imminent exploitation pressure, consistent with the low EPSS score (0.313%). The realistic risk profile is closer to "moderate" than the raw CVSS suggests: it's a data-integrity/sabotage primitive for insiders or attackers who have already obtained a foothold with notebook-edit rights, not a remote initial-access vector.

How does the attack unfold?

Initial Access

Attacker obtains or already holds valid low-privilege credentials with the NOTEBOOK_EDIT permission on the Velociraptor deployment.

AML.T0012

Exploitation

Attacker sends a crafted NewNotebook API request with a path-traversal payload in the notebook parameter, satisfying the required ".json.db" extension while escaping the org's data store directory.

Impact

The write overwrites an existing ACL or hunt metadata file elsewhere on disk, corrupting access-control state or hunt records and degrading platform integrity and investigative reliability.

Initial Access

Attacker obtains or already holds valid low-privilege credentials with the NOTEBOOK_EDIT permission on the Velociraptor deployment.

AML.T0012

Exploitation

Attacker sends a crafted NewNotebook API request with a path-traversal payload in the notebook parameter, satisfying the required ".json.db" extension while escaping the org's data store directory.

Impact

The write overwrites an existing ACL or hunt metadata file elsewhere on disk, corrupting access-control state or hunt records and degrading platform integrity and investigative reliability.

How severe is it?

CVSS 3.1

7.1 / 10

EPSS

0.4%

chance of exploitation in 30 days

Higher than 38% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

No known exploitation

Sophistication

Moderate

What is the attack surface?

AVNetwork

ACLow

PRLow

UINone

SUnchanged

CNone

IHigh

ALow

What should I do?

1 step

Apply the vendor-supplied patch referenced in the Velociraptor advisory (docs.velociraptor.app/announcements/advisories/cve-2026-18640/) as soon as it is validated in a test environment. Until patched, restrict the NOTEBOOK_EDIT permission to a minimal set of trusted, audited accounts, since possession of that permission is the sole prerequisite for exploitation. Add file-integrity monitoring on the org data store directory (particularly ACL and hunt record files) to detect unexpected writes or unauthorized ".json.db" files appearing outside expected notebook paths, and review Velociraptor audit logs for NewNotebook API calls containing path traversal sequences (e.g., "../").

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?

Auth BypassDoSAPI

Which compliance frameworks are affected?

This CVE is relevant to:

ISO 42001

A.9-Data for AI systems

Frequently Asked Questions

What is CVE-2026-18640?

A flaw in Velociraptor's NewNotebook API fails to properly sanitize file path parameters, letting any authenticated user who already holds the NOTEBOOK_EDIT permission write files outside the organization's designated data store directory. The forced ".json.db" extension limits the blast radius to metadata-type files, but that still includes access control lists and hunt records, so a malicious or compromised low-privilege user can corrupt the very files that govern who can see and do what inside the platform. This is not a remote-code-execution bug and there is no public exploit, no Nuclei template, and it is not in CISA KEV — CISA's own SSVC decision rates it TRACK, its lowest urgency tier, and EPSS puts real-world exploitation probability at just 0.3%. Because Velociraptor is frequently used as forensic and incident-response tooling, integrity damage to ACL or hunt metadata could quietly degrade an organization's ability to trust or reconstruct an active investigation, which raises the stakes above what the CVSS 7.1 score alone conveys. Patch to the fixed Velociraptor release per the vendor advisory, and in the meantime restrict NOTEBOOK_EDIT to trusted operators and monitor for unexpected writes to ACL/hunt metadata files as a detection signal.

Is CVE-2026-18640 actively exploited?

No confirmed active exploitation of CVE-2026-18640 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-18640?

Apply the vendor-supplied patch referenced in the Velociraptor advisory (docs.velociraptor.app/announcements/advisories/cve-2026-18640/) as soon as it is validated in a test environment. Until patched, restrict the NOTEBOOK_EDIT permission to a minimal set of trusted, audited accounts, since possession of that permission is the sole prerequisite for exploitation. Add file-integrity monitoring on the org data store directory (particularly ACL and hunt record files) to detect unexpected writes or unauthorized ".json.db" files appearing outside expected notebook paths, and review Velociraptor audit logs for NewNotebook API calls containing path traversal sequences (e.g., "../").

What is the CVSS score for CVE-2026-18640?

CVE-2026-18640 has a CVSS v3.1 base score of 7.1 (HIGH). The EPSS exploitation probability is 0.43%.

What is the AI security impact?

Compliance Controls Affected

ISO 42001: A.9

What are the technical details?

Original Advisory

The NewNotebook API does not sufficiently sanitize its parameters allowing an authenticated user with NOTEBOOK_EDIT permission to write the notebook record outside the org's data store directory. The file written must have an extension of ".json.db" but can otherwise overwrite other metadata files (such as ACL records, hunts etc). This can corrupt these files and cause data corruption.

Exploitation Scenario

An attacker who has obtained valid low-privilege credentials for a Velociraptor deployment — for example, an analyst account compromised via phishing or credential reuse — authenticates and issues a NewNotebook API request where the notebook identifier/name parameter contains directory traversal sequences. Because the API only enforces the ".json.db" file extension and does not otherwise constrain the resolved path, the write escapes the intended per-org notebook directory and lands on an existing ACL or hunt metadata file elsewhere in the data store, overwriting its contents. The resulting corruption can silently break access-control enforcement or disrupt active/historical hunt records, degrading the integrity of the platform's security operations and potentially masking the attacker's own subsequent activity.

Weaknesses (CWE)

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