CVE-2026-73609

Executive Summary

CVE-2026-73609 is a missing-authorization (CWE-862) information disclosure vulnerability in SiYuan, the privacy-first note-taking and knowledge-management platform developed by Thomas Eichmann. All SiYuan versions prior to v3.7.4 are affected: the getBookmarkLabels endpoint returns the complete workspace-wide bookmark label vocabulary without enforcing publish-access filtering, so anonymous visitors and publish-mode readers can enumerate labels associated with documents they cannot otherwise read. The impact is a low-severity confidentiality breach (CVSS 5.8) with no integrity or availability impact, but bookmark metadata frequently discloses sensitive subject matter and organizational structure across the entire workspace.

Technical Analysis

The vulnerability resides in the API layer of SiYuan instances exposed over HTTP(S). Specifically, the getBookmarkLabels endpoint is responsible for returning the set of bookmark labels (user-assigned tags) saved across the workspace. The NVD description states that versions prior to v3.7.4 return "all bookmark labels in the workspace without publish-access filtering." This is a textbook CWE-862 (Missing Authorization) flaw: the endpoint performs no access-control check against the requesting principal's permissions before serving the full label set.

Two attacker profiles are explicitly identified in the description:

1. Anonymous readers — unauthenticated users who reach the instance's web interface or API.
2. Publish-mode readers — users granted limited "publish" access, intended to view only selected published content.

Both profiles can successfully invoke the endpoint and receive the aggregate bookmark vocabulary. Because bookmark labels are workspace-global metadata, the response is not scoped to the requester's authorized documents. Even when document content is protected by SiYuan's access-control mechanisms, the label list acts as a high-signal metadata leak: labels frequently encode project codenames, client names, personnel identifiers, medical or legal topics, and internal organizational taxonomy.

The attack surface is any network-reachable SiYuan instance. The endpoint requires no authentication (PR:N), is trivially reachable (AC:L), and does not require user interaction. The CVSS scope is "Changed" (S:C), meaning the vulnerable component (the endpoint handler) impacts a resource beyond its security scope — namely, the entire workspace's bookmark metadata. Confidentiality impact is rated Low (C:L) because the exposure is limited to metadata rather than document bodies, although the intelligence value of that metadata can be substantial.

Insufficient technical detail is available in the NVD record to confirm the exact HTTP method, request parameters, or response serialization format of the endpoint. No code-level walkthrough, patch diff, or vendor advisory was included in the provided intelligence data, and no public PoC or exploit code is known at the time of writing.

Exploitation Scenarios

No public PoC or exploit code is available; the attack chain below is inferred strictly from the vulnerability class (CWE-862) and the endpoint behavior described in the NVD record.

1. Reconnaissance / target identification. An attacker identifies internet-exposed SiYuan instances using lightweight scanning — for example, Shodan queries such as title:SiYuan or banner/service fingerprinting. The attacker does not need valid credentials, because the vulnerable condition is reachable without authentication.

2. Vulnerability trigger. The attacker sends an HTTP request to the getBookmarkLabels endpoint on the exposed instance. Because the endpoint lacks publish-access filtering, the request is processed and a response is generated regardless of whether the requester is anonymous or has only publish-mode privileges. No authentication bypass technique, payload, or special encoding is required — the flaw is the absence of an authorization check, not a logic bug that needs to be stimulated.

3. Execution / retrieval. The response body contains the complete set of bookmark labels used across the workspace. The attacker records the label vocabulary. Since this is a pure read operation performed by the legitimate application logic, there is no malicious input, injection payload, or post-exploitation code execution.

4. Post-exploitation impact. The harvested labels are used to map the workspace's subject matter and internal organization. Even if the underlying notes remain inaccessible, the label set reveals that documents on specific topics (e.g., "merger-acme", "GDPR-breach", "HR-2026-layoffs", "patient-X") exist in the knowledge base. This enables targeted social engineering, prioritization of further attacks against document-access endpoints, and competitive intelligence gathering. In a publish-mode scenario, a user who was intentionally given limited exposure can silently escalate their visibility over workspace metadata.

Impact Assessment

From a business perspective, the primary impact is a confidentiality breach of workspace metadata. Enterprises and professional users frequently deploy SiYuan as a self-hosted knowledge base for sensitive internal material — legal matters, HR cases, patient data, proprietary product roadmaps, and client relationships. Bookmark labels are typically short, memorable, and descriptive; their disclosure is equivalent to leaking a table of contents of the organization's private notes.

CVSS 3.1 subscore analysis:

• Attack Vector: Network (AV:N) — exploitable remotely over HTTP(S).
• Attack Complexity: Low (AC:L) — no special conditions.
• Privileges Required: None (PR:N) — anonymous exploitation possible.
• User Interaction: None (UI:N).
• Scope: Changed (S:C) — the vulnerable component is the API endpoint, while the affected resource is workspace data beyond the component's authority boundary.
• Confidentiality: Low (C:L) — metadata only, not document bodies.
• Integrity / Availability: None (I:N / A:N).

The EPSS score of 0.00194 (9.4th percentile) indicates a low modeled probability of exploitation in the wild at present. CISA KEV has not listed this CVE, and no evidence of active threat-actor campaigns was present in the provided intelligence data. Nevertheless, the low exploitation barrier (unauthenticated, network-accessible, trivial request) means the risk calculus changes quickly once the vulnerability becomes widely known; the lack of a tracked patch amplifies that risk.

The blast radius covers all workspaces hosted on an affected instance, including multiple tenants or departments if SiYuan is deployed as a shared internal service. Additionally, because the disclosure is global to the workspace, a single compromise of one publish-mode reader account is sufficient to enumerate the entire organization's label vocabulary.

Detection Guidance

Sigma Rule

title: SiYuan getBookmarkLabels Endpoint Access - CVE-2026-73609
id: 9f3c1b2e-4c5d-4e6f-8a7b-1d2e3f4a5b6c
status: experimental
description: >
  Detects HTTP requests to the SiYuan getBookmarkLabels endpoint, which
  lacks publish-access filtering and exposes workspace-wide bookmark labels
  to anonymous and publish-mode readers (CVE-2026-73609, CWE-862).
author: ValtersIT.com Threat Intelligence
date: 2026/07/01
references:
  - https://nvd.nist.gov/vuln/detail/CVE-2026-73609
logsource:
  category: webserver
  product: generic
detection:
  selection:
    cs-uri-path|contains: "getBookmarkLabels"
  filter_known:
    cs-username: "admin"  # Adjust to local privileged accounts
  condition: selection and not filter_known
falsepositives:
  - Legitimate authenticated retrieval of bookmark labels by authorized administrators
level: medium
tags:
  - attack.t1190
  - cve.2026-73609
  - cwe.862

YARA Rule

rule siyuan_cve_2026_73609_getBookmarkLabels
{
    meta:
        author = "ValtersIT.com Threat Intelligence"
        description = "Detects HTTP traffic captures, proxy logs, or memory artifacts referencing the SiYuan getBookmarkLabels endpoint associated with CVE-2026-73609 information disclosure."
        reference = "https://nvd.nist.gov/vuln/detail/CVE-2026-73609"
        cve = "2026-73609"
        cwe = "862"
        date = "2026-07-01"
    strings:
        $endpoint = "getBookmarkLabels" ascii wide
        $product  = "SiYuan" ascii wide
        $label    = "bookmarkLabels" ascii wide
    condition:
        $endpoint and (#label >= 1 or $product)
}

Suricata IDS Rule

alert http any any -> any any (msg:"ET WEB_SPECIFIC_APPS SiYuan getBookmarkLabels Unauthenticated Access Attempt (CVE-2026-73609)"; flow:to_server,established; content:"getBookmarkLabels"; http.uri; nocase; classtype:attempted-recon; sid:2026736091; rev:1; metadata:created_at 2026_07_01, updated_at 2026_07_01, cve [2026-73609];)

# Complementary signature for the server response (label list exfiltration):
alert http any any -> any any (msg:"ET WEB_SPECIFIC_APPS SiYuan getBookmarkLabels Response with Workspace Labels (CVE-2026-73609)"; flow:to_client,established; content:"getBookmarkLabels"; http.response_uri; nocase; content:"["; http.response_body; fast_pattern; classtype:info-leak; sid:2026736092; rev:1;)

Elastic Detection

// Elastic EQL — detects HTTP requests to the vulnerable endpoint from non-local sources
network where http.request.method in ("GET", "POST") and http.request.uri like "*getBookmarkLabels*" and not source.ip in ("127.0.0.1", "::1")

Splunk SPL Query

index=* sourcetype=access_combined uri=*getBookmarkLabels*
| eval access_profile=case(user=="-" OR user=="anonymous", "ANONYMOUS", user=="publish" OR user=="publish-user", "PUBLISH_MODE", true(), "AUTHENTICATED")
| stats count earliest(_time) latest(_time) values(clientip) values(status) by uri, user, access_profile
| where access_profile in ("ANONYMOUS", "PUBLISH_MODE")
| sort - count

Wazuh OSSEC Rule

<group name="siyuan,cve_2026_73609,">
  <!-- Requires the standard web-access log decoder (31450 -> 3150 family) -->
  <rule id="403330" level="10" frequency="6" timeframe="120">
    <if_sid>3150</if_sid>
    <match>getBookmarkLabels</match>
    <description>Possible CVE-2026-73609 exploitation: repeated access to the SiYuan getBookmarkLabels endpoint (workspace bookmark label enumeration).</description>
    <group>siyuan,web_application,attack.t1190,cwe.862,</group>
  </rule>
</group>

Mitigation & Patch Guidance

Patch confidence is 0/4 (UNPATCHED) — no vendor patch evidence is available from any tracked source in the provided intelligence data. Notably, the NVD description itself identifies v3.7.4 as the version containing the fix, so the upgrade path is clear even though no formal advisory has been ingested.

1. Immediate mitigations (while unpatched):

   • Restrict network access to SiYuan web interfaces with firewall/ACL rules; do not expose the instance directly to the internet.
   • Place the instance behind a reverse proxy (e.g., nginx, Apache, Caddy) that enforces mandatory authentication (SSO, MFA) for all traffic before it reaches the SiYuan backend.
   • Disable or disable publish-mode features if external sharing is not actively required by the business.
   • If publish-mode is required, ensure only low-sensitivity content is published and restrict the publish account to a dedicated, minimal-access workspace.
   • Deploy WAF or IDS/IPS rules blocking unauthenticated requests to URIs containing getBookmarkLabels (see Suricata and Sigma rules above).
   • Audit existing access logs for prior hits to the endpoint from anonymous or publish-mode sources.

2. Patch / upgrade steps:

   • Monitor the official SiYuan distribution channels for the v3.7.4 release (or later).
   • When available, back up the workspace database and configuration, then upgrade to v3.7.4 or newer following the vendor's official upgrade procedure.
   • After upgrade, functionally test that the getBookmarkLabels endpoint now returns labels only as permitted by effective publish-access rights.
   • Re-verify patch status in threat-intelligence feeds until the patch confidence score rises above 0/4.

3. Long-term hardening:

   • Perform an access-control review of all SiYuan API endpoints; enforce authorization at the API layer, not only in the UI.
   • Adopt a principle of least privilege for all workspace roles, including publish-mode roles.
   • Educate users on bookmark-label hygiene: avoid embedding client names, personnel identifiers, medical conditions, or project codenames in labels; if possible, adopt a coding scheme for internal labels.
   • Enable centralized web-server logging and ship logs to SIEM/EDR pipelines to support the detection rules above.
   • Track this CVE in vulnerability-management systems using the CVSS 5.8 vector and EPSS 0.00194 baseline, and re-evaluate exposure as new exploit evidence emerges.

References

1. NVD Entry — https://nvd.nist.gov/vuln/detail/CVE-2026-73609
2. CWE-862 Missing Authorization — https://cwe.mitre.org/data/definitions/862.html
3. SiYuan Official GitHub Repository (version/release tracking) — https://github.com/siyuan-note/siyuan
4. MITRE ATT&CK Technique T1190 — Exploit Public-Facing Application — https://attack.mitre.org/techniques/T1190/
5. CISA Known Exploited Vulnerabilities Catalog — https://www.cisa.gov/known-exploited-vulnerabilities-catalog (CVE-2026-73609 is NOT currently listed)

Exploit-in-the-wild evidence: None available in the provided intelligence data. No Nuclei template, no exploit code, and no documented threat-actor campaign references this CVE as of the report date.