GHSA-q2vg-7qgx-x5fc CVE-2026-72811 GO-2026-6402
SiYuan: SQL injection in backlink/mention search via unescaped stored and client input (publish mode): first-order (client keyword) and second-order (stored document title) breakout on read-write handle
- Severity
- 9.9 / 10
- Critical · CVSS 4.0
- Exploit chance (EPSS)
- 0.44%
- Chance of exploitation in the next 30 days. Higher than 37.96% of scored vulnerabilities.
- Known exploited
- No
- Not in CISA's catalog
- Weakness
- CWE-89 Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection')
- Fix
0.0.0-20260723004839-1a5b3431d5ab- Upgrade github.com/siyuan-note/siyuan/kernel to this version or later
Description
CVE: This vulnerability corresponds to CVE-2026-72811.
Summary
The backlink/mention search query (kernel/model/backlink.go) concatenates stored block metadata (title, name, alias, anchor text) and the client-supplied keyword into a SQL MATCH/search statement, escaping only the double-quote character (") and not the single quote ('). A single quote in either the client keyword or in stored document metadata breaks out of the string literal. The query runs on the main read-write siyuan.db handle through a statement-stacking-capable driver.
This yields two vectors:
- First-order: a client-supplied keyword containing
'injects directly. This path is reachable by an anonymous reader on the publish surface. - Second-order: a document whose title/name/alias contains
'is stored safely (indexing uses parameterized inserts) but detonates when that stored value is later concatenated into the backlink query including on another user's kernel that has ingested the malicious document.
Details
Storage is safe; reuse is not. Indexing INSERTs (kernel/sql/upsert.go) are parameterized ((?,?,…) with bound arguments for Name/Content/Markdown/IAL), so malicious .sy content is stored intact and safely. The injection is in the reuse path: the backlink/mention MATCH query (kernel/model/backlink.go, around line 980) builds its condition by concatenating the stored title/name/alias/anchor and the client keyword, applying only "→escaping and not '. A ' in either source terminates the literal and lands in SQL context.
Sink / handle. The concatenated statement reaches SelectBlocksRawStmtNoParse → query() on the global read-write siyuan.db handle (DSN has no mode=ro/_query_only), driven by the vendored 88250/go-sqlite3 fork whose connection query loops over ;-separated statements (stacking possible). Ceiling is arbitrary SQL cross-notebook read and write.
Route / auth tier. The backlink/mention family (getBacklink, getBacklink2, getBacklinkDoc, getBackmentionDoc) is CheckAuth-only, so the first-order client-keyword vector is reachable by the publish RoleReader token and by the anonymous account when Publish.Auth.Enable is false.
Scope of the injection surface (confirmed the only second-order sink). A sweep of the query paths that reuse stored content confirmed this is the sole injectable reuse sink: indexing INSERTs are parameterized, tag search uses content LIKE ? with pattern escaping, virtual refs use in-memory matching (no SQL), IN(...) joins use node-IDs/hex-hashes/ROWIDs (no quote surface), and the graph name/content filter escapes '. Only the backlink/mention MATCH path concatenates with ' unescaped.
Proof of Concept
Reproduced on a local instance (SiYuan running locally, publish mode enabled on port 6808, publish Basic Auth disabled). Strictly non-destructive verification, a syntax-error probe and a stored-value observation; no UNION exfiltration, no write, no DDL.
First-order (client keyword), anonymous reader on port 6808:
POST http://127.0.0.1:6808/api/ref/getBacklink2
{"id":"<block id>","k":"a'b","mk":""}
The keyword a'b reaches the concatenated MATCH condition and produces a SQL parse error, confirming the single quote breaks out of the literal and the client keyword lands in SQL context.
Second-order (stored title), setup via admin then observed anonymously:
A document whose title contains a single quote e.g. ("locked-doc") style metadata is stored safely by the parameterized indexer, then appears at the exact unescaped position in the backlink query when that document participates in a backlink/mention lookup, breaking the query the same way. This demonstrates that stored document metadata detonates on reuse, independent of the client keyword.
Verification was limited to the syntax-error probe and the stored-value position observation; no exfiltration or write statement was executed. Full reproduction detail available privately on request.
Impact
First-order: an anonymous reader (publish mode with auth disabled) or any publish RoleReader injects arbitrary SQL into the backlink/mention query via the keyword parameter, on the read-write main handle: cross-notebook read disclosure and, via statement stacking, write and ATTACH. This is the same anonymous-arbitrary-SQL severity as the search/embed injection findings, on a different sink.
Second-order: a document whose title/name/alias contains injection content is stored safely but executes when that stored value is concatenated into the backlink query. This fires on any kernel that has ingested the malicious document for example via a shared or imported .sy file, sync, or an opened .sy.zip. The precondition here is content delivery into the victim workspace (not a direct anonymous request), so it is a distinct, delivery-dependent threat model from the first-order vector.
Encrypted notebooks are out of scope; code execution is not reachable in the default build (no load_extension).
Distinctness from the other SQL findings: this is a different sink (backlink.go MATCH query) from the searchDocs (LIKE keyword), searchEmbedBlock (verbatim stmt), and getEmbedBlock (IN-clause ID array) findings, and it is the only one with a second-order (stored-content) vector. A fix to any of those does not remediate this path.
Suggested fix
Parameterize the backlink/mention query bind the keyword and any stored metadata as placeholders rather than concatenating or escape ' (and use LIKE-pattern escaping) consistently, matching the tag-search and graph-filter paths that already do so. The second-order vector specifically requires that stored metadata be bound rather than concatenated, since escaping at storage time is not the defense (storage is already safe); the fix must be at the query-construction site.
Affected versions
Go
| Package | Vulnerable | Fixed in |
|---|---|---|
| github.com/siyuan-note/siyuan/kernel | < 0.0.0-20260723004839-1a5b3431d5ab |
0.0.0-20260723004839-1a5b3431d5ab |
How to fix
Move to the fixed version or later with your package manager, update the lockfile, and run your tests. If the package only arrives through another dependency, upgrade that dependency first. If none of its releases pulls a fixed version yet, force the version with your package manager's override.
github.com/siyuan-note/siyuan/kernel Go → 0.0.0-20260723004839-1a5b3431d5ab
- go
go get github.com/siyuan-note/siyuan/kernel@v0.0.0-20260723004839-1a5b3431d5ab && go mod tidy
Severity breakdown
CVSS 4.0 9.9
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:N/SC:H/SI:H/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X
- Attack vector
- Network
- Attack complexity
- Low
- Attack requirements
- None
- Privileges required
- None
- User interaction
- None
- Confidentiality (vulnerable system)
- High
- Integrity (vulnerable system)
- High
- Availability (vulnerable system)
- None
- Confidentiality (subsequent systems)
- High
- Integrity (subsequent systems)
- High
- Availability (subsequent systems)
- None
- Exploit maturity
- Not defined
- Scope
- X
CVSS 3.1 10.0
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:N
- Attack vector
- Network
- Attack complexity
- Low
- Privileges required
- None
- User interaction
- None
- Scope
- Changed
- Confidentiality
- High
- Integrity
- High
- Availability
- None
Weakness
CWE-89: Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection')
The product constructs all or part of an SQL command using externally-influenced input from an upstream component, but it does not neutralize or incorrectly neutralizes special elements that could modify the intended SQL command when it is sent to a downstream component. Without sufficient removal or quoting of SQL syntax in user-controllable inputs, the generated SQL query can cause those inputs to be interpreted as SQL instead of ordinary user data.
How to prevent it
- Architecture and Design: Use a vetted library or framework that does not allow this weakness to occur or provides constructs that make this weakness easier to avoid [REF-1482]. For example, consider using persistence layers such as Hibernate or Enterprise Java Beans, which can provide significant protection against SQL injection if used properly.
- Architecture and Design: If available, use structured mechanisms that automatically enforce the separation between data and code. These mechanisms may be able to provide the relevant quoting, encoding, and validation automatically, instead of relying on the developer to provide this capability at every point where output is generated. Process SQL queries using prepared statements, parameterized queries, or stored procedures. These features should accept parameters or variables and support strong typing. Do not dynamically construct and execute query strings within these features using "exec" or similar functionality, since this may re-introduce the possibility of SQL injection. [REF-867]
- Architecture and Design, Operation: Run your code using the lowest privileges that are required to accomplish the necessary tasks [REF-76]. If possible, create isolated accounts with limited privileges that are only used for a single task. That way, a successful attack will not immediately give the attacker access to the rest of the software or its environment. For example, database applications rarely need to run as the database administrator, especially in day-to-day operations. Specifically, follow the principle of least privilege when creating user accounts to a SQL database. The database users should only have the minimum privileges necessary to use their account. If the requirements of the system indicate that a user can read and modify their own data, then limit their privileges so they cannot read/write others' data. Use the strictest permissions possible on all database objects, such as execute-only for stored procedures.
- 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.
- Implementation: While it is risky to use dynamically-generated query strings, code, or commands that mix control and data together, sometimes it may be unavoidable. Properly quote arguments and escape any special characters within those arguments. The most conservative approach is to escape or filter all characters that do not pass an extremely strict allowlist (such as everything that is not alphanumeric or white space). If some special characters are still needed, such as white space, wrap each argument in quotes after the escaping/filtering step. Be careful of argument injection (CWE-88). Instead of building a new implementation, such features may be available in the database or programming language. For example, the Oracle DBMS_ASSERT package can check or enforce that parameters have certain properties that make them less vulnerable to SQL injection. For MySQL, the mysql_real_escape_string() API function is available in both C and PHP.
- 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, denylists can be useful for detecting potential attacks or determining which inputs are so malformed that they should be rejected outright. When constructing SQL query strings, use stringent allowlists that limit the character set based on the expected value of the parameter in the request. This will indirectly limit the scope of an attack, but this technique is less important than proper output encoding and escaping. Note that proper output encoding, escaping, and quoting is the most effective solution for preventing SQL injection, although input validation may provide some defense-in-depth. This is because it effectively limits what will appear in output. Input validation will not always prevent SQL injection, especially if you are required to support free-form text fields that could contain arbitrary characters. For example, the name "O'Reilly" would likely pass the validation step, since it is a common last name in the English language. However, it cannot be directly inserted into the data
- Architecture and Design: When the set of acceptable objects, such as filenames or URLs, is limited or known, create a mapping from a set of fixed input values (such as numeric IDs) to the actual filenames or URLs, and reject all other inputs.
- Implementation: Ensure that error messages only contain minimal details that are useful to the intended audience and no one else. The messages need to strike the balance between being too cryptic (which can confuse users) or being too detailed (which may reveal more than intended). The messages should not reveal the methods that were used to determine the error. Attackers can use detailed information to refine or optimize their original attack, thereby increasing their chances of success. If errors must be captured in some detail, record them in log messages, but consider what could occur if the log messages can be viewed by attackers. Highly sensitive information such as passwords should never be saved to log files. Avoid inconsistent messaging that might accidentally tip off an attacker about internal state, such as whether a user account exists or not. In the context of SQL Injection, error messages revealing the structure of a SQL query can help attackers tailor successful attack strings.
- Operation: Use an application firewall that can detect attacks against this weakness. It can be beneficial in cases in which the code cannot be fixed (because it is controlled by a third party), as an emergency prevention measure while more comprehensive software assurance measures are applied, or to provide defense in depth [REF-1481.
- Operation, Implementation: When using PHP, configure the application so that it does not use register_globals. During implementation, develop the application so that it does not rely on this feature, but be wary of implementing a register_globals emulation that is subject to weaknesses such as CWE-95, CWE-621, and similar issues.
References
Advisories
Packages
Other links
Timeline
- Published
- Last updated by the source
- Latest EPSS score
Related advisories
| Advisory | Severity | Why it's related |
|---|---|---|
| GHSA-xp3q-r38w-vgqm SiYuan: The publish filter for local storage returns the administrator's entire storage map with only three keys sanitized | Medium | Also affects github.com/siyuan-note/siyuan/kernel |
| GHSA-9cqf-hhrq-7v45 SiYuan: getAttributeViewSearchTarget returns database row content to anonymous readers with no publish-access check, reopening the class closed one day earlier at the adjacent route | High | Also affects github.com/siyuan-note/siyuan/kernel |
| GHSA-57v5-wqx3-cgj4 SiYuan: Database view structure (all view names, layout types and per-field visibility) is returned to anonymous readers by /api/av/getAttributeViewFieldViews | Medium | Also affects github.com/siyuan-note/siyuan/kernel |
| GHSA-4wwp-f6gw-6qm5 SiYuan: TLS Private Keys Readable via getFile (Incomplete Blocklist) | High | Also affects github.com/siyuan-note/siyuan/kernel |
| GHSA-3cm4-ccvw-6xr6 SiYuan: /history/*path and /repo/diff/*path potentially exposing historical snapshots of data/.siyuan/publishAccess.json and data/templates/* | Medium | Also affects github.com/siyuan-note/siyuan/kernel |
| GHSA-x8gv-g2g3-65fj SiYuan Agent Tools SSRF via DNS-Rebinding TOCTOU (Bypass of CheckHostSSRF) | High | Also affects github.com/siyuan-note/siyuan/kernel |
| GHSA-p23f-cm6q-2qp8 SiYuan MCP asset.upload Reads Arbitrary Absolute File Paths (Workspace Boundary Bypass) | Medium | Also affects github.com/siyuan-note/siyuan/kernel |
| GHSA-rpx2-p6hp-x5gj SiYuan: Missing consistency check between packageName and repoURL/repoHash in bazaar package install allows overwriting an existing, trusted plugin | Medium | Also affects github.com/siyuan-note/siyuan/kernel |
MergeHarbor checks every repository's dependencies against this database, ranks what matters by exploitability and reachability, routes it to the owning team and opens the fix pull request.
Check your repos with MergeHarbor