CVE-2026-73609: SiYuan: missing authz leaks bookmark labels in publish mode

GHSA-j4ph-9xwf-wcj4MEDIUMPoC AVAILABLE
Published October 1, 2026
CISO Take

SiYuan's /api/attr/getBookmarkLabels endpoint only checks authentication and applies no publish-access filtering, so an anonymous or reader-role user in publish mode can retrieve every bookmark label in the workspace, including labels from hidden, password-protected or forbidden documents. Impact is limited to label text (no block IDs or content), but labels often name projects, people and subject matter, so this is a real confidentiality leak for teams using SiYuan as a personal or AI-assisted knowledge base. Exploitation is trivial, a single unauthenticated POST, though the EPSS score of 0.00325 (top 76th percentile) is low, there is no public exploit or Nuclei template, and it is not in CISA KEV. With 3,047 downstream dependents and 62 other CVEs in the same package, the exposure is real but contained. Upgrade to SiYuan v3.8.0 (kernel patched at 0.0.0-20260812083335-251596fc0de2), and until then disable publish mode or put it behind authenticated access.

Sources:NVDGitHub AdvisoryEPSSATLASOpenSSFvulncheck.com

What is the risk?

MEDIUM (CVSS 5.8, scope changed, confidentiality low only). Attack is network-reachable, requires no privileges or user interaction and has trivial complexity. Likelihood of exploitation is low-to-moderate per EPSS (0.00325, top 76%), with no public exploit, no scanner template and no KEV listing. Risk is concentrated in instances exposing publish mode to the internet without auth. There is no integrity or availability impact. The package has a modest OpenSSF Scorecard (5.8/10) and a history of 62 other CVEs, which signals recurring authorization gaps in the publish layer.

How does the attack unfold?

Discovery
Attacker identifies a SiYuan instance with publish mode enabled and anonymous or reader access on port 6808.
AML.T0006
Exploitation
Attacker sends an empty POST to /api/attr/getBookmarkLabels, which passes CheckAuth without any publish-access filter.
AML.T0049
Collection
The server returns every distinct bookmark label in the workspace, including those from hidden, password-protected or forbidden documents.
AML.T0036
Impact
Attacker learns project names, personal and organisational names and work themes, enabling targeted follow-on attacks.

What systems are affected?

PackageEcosystemVulnerable RangePatched
Jupyter Notebookgo< 0.0.0-20260812083335-251596fc0de20.0.0-20260812083335-251596fc0de2
13.5KOpenSSF 5.73.0K dependentsPushed 7d ago80% patched~74d to patchFull package profile →

Do you use Jupyter Notebook? You're affected.

How severe is it?

CVSS 3.1
5.8 / 10
EPSS
0.3%
chance of exploitation in 30 days
Higher than 26% of all CVEs
Exploitation Status
Exploit Available
Exploitation: MEDIUM
Sophistication
Trivial
Exploitation Confidence
medium
○Public PoC indexed (trickest/cve)
Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AVACPRUISCIA
AVNetwork
ACLow
PRNone
UINone
SChanged
CLow
INone
ANone

What should I do?

1 step
  1. 1) Upgrade SiYuan to v3.8.0 or later (kernel version 0.0.0-20260812083335-251596fc0de2 or newer). 2) Until patched, disable publish mode or restrict it to authenticated readers (Publish.Auth.Enable=true) and limit access with network controls or a reverse proxy. 3) Block or monitor POST requests to /api/attr/getBookmarkLabels from publish/reader sessions at the proxy or WAF. 4) Review bookmark labels for sensitive wording and avoid placing confidential names in labels on published workspaces. 5) Check access logs for unexpected calls to this endpoint from anonymous or unknown clients.

What does CISA's SSVC say?

DecisionTrack
Exploitationnone
AutomatableYes
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:

ISO 42001
A.7.4-Quality and protection of data for AI systems
NIST AI RMF
MANAGE 2.3-Risk treatment and response to previously unknown risks
OWASP LLM Top 10
LLM02-Sensitive Information Disclosure

Frequently Asked Questions

What is CVE-2026-73609?

SiYuan's /api/attr/getBookmarkLabels endpoint only checks authentication and applies no publish-access filtering, so an anonymous or reader-role user in publish mode can retrieve every bookmark label in the workspace, including labels from hidden, password-protected or forbidden documents. Impact is limited to label text (no block IDs or content), but labels often name projects, people and subject matter, so this is a real confidentiality leak for teams using SiYuan as a personal or AI-assisted knowledge base. Exploitation is trivial, a single unauthenticated POST, though the EPSS score of 0.00325 (top 76th percentile) is low, there is no public exploit or Nuclei template, and it is not in CISA KEV. With 3,047 downstream dependents and 62 other CVEs in the same package, the exposure is real but contained. Upgrade to SiYuan v3.8.0 (kernel patched at 0.0.0-20260812083335-251596fc0de2), and until then disable publish mode or put it behind authenticated access.

Is CVE-2026-73609 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2026-73609, increasing the risk of exploitation.

How to fix CVE-2026-73609?

1) Upgrade SiYuan to v3.8.0 or later (kernel version 0.0.0-20260812083335-251596fc0de2 or newer). 2) Until patched, disable publish mode or restrict it to authenticated readers (Publish.Auth.Enable=true) and limit access with network controls or a reverse proxy. 3) Block or monitor POST requests to /api/attr/getBookmarkLabels from publish/reader sessions at the proxy or WAF. 4) Review bookmark labels for sensitive wording and avoid placing confidential names in labels on published workspaces. 5) Check access logs for unexpected calls to this endpoint from anonymous or unknown clients.

What systems are affected by CVE-2026-73609?

This vulnerability affects the following AI/ML architecture patterns: Knowledge base / note-taking platforms, Self-hosted publish services, RAG source knowledge stores.

What is the CVSS score for CVE-2026-73609?

CVE-2026-73609 has a CVSS v3.1 base score of 5.8 (MEDIUM). The EPSS exploitation probability is 0.33%.

What is the AI security impact?

Affected AI Architectures

Knowledge base / note-taking platformsSelf-hosted publish servicesRAG source knowledge stores

MITRE ATLAS Techniques

AML.T0036Data from Information Repositories
AML.T0049Exploit Public-Facing Application

Compliance Controls Affected

ISO 42001: A.7.4
NIST AI RMF: MANAGE 2.3
OWASP LLM Top 10: LLM02

What are the technical details?

Original Advisory

### Summary `/api/attr/getBookmarkLabels` is registered with `CheckAuth` only and applies no filtering of any kind. It runs a scan of the entire `blocks` table and returns the distinct set of every bookmark label in the workspace. An anonymous reader in publish mode receives the author's complete bookmark vocabulary, regardless of whether the bookmarked blocks live in published, hidden, password-protected or forbidden documents. The adjacent endpoint that returns bookmarks with their blocks does filter, and does so in a way that makes the intended rule explicit: it drops a label entirely when no accessible block carries it. ### Details **Route.** `kernel/api/router.go:297` on master, `:300` on the development branch: ```go ginServer.Handle("POST", "/api/attr/getBookmarkLabels", model.CheckAuth, getBookmarkLabels) ``` No `CheckAdminRole`, no `CheckReadonly`. Reachable by the publish `RoleReader` token and anonymously when `Publish.Auth.Enable` is `false`. **The handler** (`kernel/api/attr.go`) is a single line: ```go ret.Data = model.BookmarkLabels() ``` which reaches `kernel/model/bookmark.go:184` and then `sql.QueryBookmarkLabels()` at `kernel/sql/block_query.go:315`: ```go sqlStmt := "SELECT * FROM blocks WHERE ial LIKE ?" // "%bookmark=%" // collect distinct ialAttr(block.IAL, "bookmark") into a set, sort, return ``` There is no notebook scoping, no path scoping, no publish-access check and no filter function anywhere on the path. Every block in the workspace carrying a bookmark attribute contributes its label to the result. Bookmark labels are user-authored organisational text. They routinely name the thing being organised, which is the reason they are useful and also the reason they disclose. **The guarded sibling states the rule outright.** `getBookmark` (`kernel/api/router.go:190` on master, `:193` on the development branch) carries the same `CheckAuth`-only exposure and handles the same data. It does this: ```go bookmark.Blocks = model.FilterBlocksByPublishAccess(c, publishAccess, bookmark.Blocks) bookmark.Count = len(bookmark.Blocks) if bookmark.Count > 0 { *tempBookmarks = append(*tempBookmarks, bookmark) } ``` The final condition is the significant part. The label is not merely returned with an empty block list, it is omitted from the response entirely when no accessible block carries it. The project has therefore already decided that a reader must not learn a bookmark label they have no accessible block for. `getBookmarkLabels` returns those same labels unconditionally, one endpoint away. **Relationship to the tag-label fix.** Commit `4515fa257` ("Enforce publish access for tag labels") established that user-authored organisational labels are data the publish boundary must protect. Bookmark labels are the same class of data, drawn from the same IAL attribute store, reached through a structurally similar handler. That commit touched `kernel/api/tag.go`, `kernel/model/publish_access.go` and its test file, and nothing in `attr.go` or the bookmark path, so this endpoint was not covered by it. Worth noting that the defect here is broader than the one in the tag path. The tag filter existed and checked the visible tier, so it correctly excluded forbidden documents and leaked only the password tier. `getBookmarkLabels` has no filter at all, so labels from forbidden documents are returned as well. ### Proof of Concept Precondition: publish mode enabled (default port 6808), anonymous when `Publish.Auth.Enable` is `false`, otherwise any publish reader account. At least one bookmarked block in a document that is not published. ``` POST http://127.0.0.1:6808/api/attr/getBookmarkLabels {} → 200, a sorted array of every distinct bookmark label in the workspace, including labels used only inside documents the reader has no publish access to ``` No arguments are required. For contrast, the sibling endpoint returns nothing for those same labels to the same session: ``` POST http://127.0.0.1:6808/api/attr/getBookmark {} → 200, labels omitted entirely where no accessible block carries them ``` One endpoint withholds the label, the other returns it. ### Impact An anonymous reader in publish mode, or any publish `RoleReader`, obtains the author's complete bookmark vocabulary across the workspace. Bookmark labels are free text chosen to describe what is being organised, so the set discloses subject matter, project names, personal and organisational names, and the general shape of what the author is working on, including material in documents that are hidden, password-protected or explicitly forbidden to readers. Scoped honestly: this returns labels only. No block identifiers, document identifiers or content are included, so it does not extend into the identifier-to-content resolution paths. Confidentiality only, with no integrity or availability impact. ### Suggested fix Mirror `getBookmark`. For read-only roles, build the label set from blocks that have passed `FilterBlocksByPublishAccess`, and emit a label only if at least one surviving block carries it. That is the rule the sibling endpoint already implements, applied to the same data.

Exploitation Scenario

An adversary finds a SiYuan instance exposing publish mode on port 6808 with anonymous access. They send an empty POST to /api/attr/getBookmarkLabels and receive a sorted list of all bookmark labels, including ones such as client names, internal project codenames or incident topics tied to unpublished documents. They use this to map the author's work, to craft targeted phishing or social engineering, or to identify what valuable material to pursue through other flaws in the same package.

Weaknesses (CWE)

CWE-862 — Missing Authorization: The product does not perform an authorization check when an actor attempts to access a resource or perform an action.

  • [Architecture and Design] Divide the product into anonymous, normal, privileged, and administrative areas. Reduce the attack surface by carefully mapping roles with data and functionality. Use role-based access control (RBAC) [REF-229] to enforce the roles at the appropriate boundaries. Note that this approach may not protect against horizontal authorization, i.e., it will not protect a user from attacking others with the same role.
  • [Architecture and Design] Ensure that access control checks are performed related to the business logic. These checks may be different than the access control checks that are applied to more generic resources such as files, connections, processes, memory, and database records. For example, a database may restrict access for medical records to a specific database user, but each record might only be intended to be accessible to the patient and the patient's doctor [REF-7].

Source: MITRE CWE corpus.

CVSS Vector

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

Timeline

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

Related Vulnerabilities