CVEReports
CVEReports

Automated vulnerability intelligence platform. Comprehensive reports for high-severity CVEs generated by AI.

Product

  • Home
  • Sitemap
  • RSS Feed

Company

  • About
  • Contact
  • Privacy Policy
  • Terms of Service

© 2026 CVEReports. All rights reserved.

Made with love by Amit Schendel & Alon Barad



CVE-2026-22028

Preact's Identity Crisis: When JSON Becomes Code via CVE-2026-22028

Amit Schendel
Amit Schendel
Senior Security Researcher

Feb 24, 2026·6 min read·79 visits

Executive Summary (TL;DR)

Preact forgot how to tell the difference between a real Virtual DOM node and a JSON object masquerading as one. If you pass user-controlled JSON into a render tree, attackers can inject arbitrary HTML components and execute JavaScript.

A high-severity type confusion vulnerability in Preact allows attackers to smuggle malicious Virtual DOM nodes via simple JSON payloads, leading to Cross-Site Scripting (XSS). A regression in version 10.26.5 weakened the internal validation logic, causing the framework to mistake plain JavaScript objects for trusted VNodes.

The Hook: Trust Issues in the Virtual DOM

In the modern web ecosystem, we've largely outsourced our DOM manipulation anxiety to frameworks like React and Preact. We trust them to take our messy state and efficiently paint the screen. We assume that when we write <div>{userInput}</div>, the framework will treat userInput as exactly that: text. Maybe a number. But certainly not executable code.

Preact, the lightweight cousin of React, operates on Virtual DOM (VNode) objects. These are lightweight JavaScript objects that describe what the UI should look like. To prevent chaos, Preact (and React) historically rely on a mechanism to distinguish a "blessed" VNode created by the framework from a random JSON object sent by an API.

Usually, this is done via Symbols or non-enumerable properties—things that JSON.stringify() destroys and JSON.parse() cannot recreate. It is the digital equivalent of a holographic watermark on a banknote. If the watermark isn't there, it's just monopoly money.

But in CVE-2026-22028, Preact stopped checking for the watermark. It started accepting the monopoly money. And unfortunately for us, the attackers are printing cash.

The Flaw: A Case of Mistaken Identity

The vulnerability is a classic Type Confusion bug, specifically CWE-843. The core issue lies in how Preact decides if a JavaScript object is a valid VNode that should be rendered as an element, or just a plain object that should be stringified or ignored.

Prior to version 10.26.5, Preact enforced strict checks. It looked for specific internal markers—often Symbols or specific integer constants—that an attacker cannot inject via a standard JSON payload because JSON doesn't support them. You can't put Symbol('react.element') in a JSON file. It just doesn't parse.

However, a regression introduced in 10.26.5 "softened" these checks. The developers likely tried to optimize the hot path of the renderer or support a specific edge case for serialization. In doing so, they inadvertently allowed "Duck Typing" on VNodes.

If it looks like a VNode (has a props property, has a type property) and quacks like a VNode (has a __v property set to 0 or null), Preact says, "Come on in!" This allows an attacker to construct a JSON object that satisfies these loose constraints, effectively forging a VNode without ever calling createElement.

The Code: Removing the Safety Rails

Let's look at the logic. In a secure implementation, the check often involves a Symbol. Since JSON.parse returns "Plain Old JavaScript Objects" (POJOs), a secure check fails immediately against network data.

The Secure Pattern (Conceptual):

// Inside Preact's render loop
if (child && child.$$typeof === REACT_ELEMENT_SYMBOL) {
  // It's a real VNode, render it.
  renderVNode(child);
} else {
  // It's just data. Escape it and show text.
  document.createTextNode(String(child));
}

The Vulnerable Pattern (Regression):

Around version 10.26.5, the logic shifted to something that looks dangerously like this:

// The check became looser to support some internal optimization
if (child && (child.__v === 0 || child.constructor === undefined)) {
   // "Looks legit to me!"
   renderVNode(child);
}

The critical failure here is that __v: 0 is perfectly valid JSON. An attacker doesn't need to bypass a memory protection or guess a randomization seed; they just need to send { "__v": 0 } in their payload. The framework creates a VNode instance from this raw data, essentially letting the attacker define the type (HTML tag) and props (attributes) of the rendered element.

The Exploit: Weaponizing JSON

How does a hacker turn this into a weapon? Imagine a user profile page that fetches data from an API. The frontend code looks innocent enough:

// Vulnerable component
const Profile = ({ userBio }) => {
  // userBio comes directly from JSON.parse(apiResponse)
  return <div>{userBio}</div>;
};

Normally, if userBio is a string, it renders text. If userBio is an object, React/Preact usually warns you or stringifies it to [object Object]. But thanks to the type confusion, we can feed it a structure that Preact interprets as a command to render an image tag with an XSS payload.

The Payload:

{
  "type": "img",
  "props": {
    "src": "invalid_url_to_trigger_error",
    "onerror": "alert('System Compromised via Preact!');"
  },
  "__v": 0
}

When the application processes this:

  1. The API returns the JSON.
  2. The application passes the object into the JSX {userBio} expression.
  3. Preact inspects the object.
  4. It sees __v: 0 and assumes this is a valid VNode.
  5. It reads type: "img" and creates an <img> DOM element.
  6. It applies the props. src fails to load, triggering onerror.
  7. Boom. Arbitrary JavaScript execution in the victim's session.

This isn't just an alert box. It's cookie theft, session hijacking, or forcing the user to perform actions (CSRF) on the platform.

The Impact: Why This Matters

This vulnerability is particularly nasty because it turns standard data handling into a remote code execution vector (in the browser context). Developers are taught to sanitize HTML strings, but they are rarely taught to sanitize JSON objects passed to the render tree because the framework is supposed to handle that isolation.

If you are using Preact 10.26.5 through 10.28.1, any endpoint where a user can store a JSON object that gets rendered into a view is a potential XSS vector. This includes:

  • User Profiles: Bio fields, custom statuses.
  • Comments: If the comment body is stored as JSON.
  • Config Dashboards: Where settings might be rendered dynamically.

The CVSS score is 7.2 (High), but in the right application—say, a chat app or a dashboard with high privileges—the impact is critical. It bypasses standard XSS filters that look for <script> tags because the payload is just a JSON object.

The Fix: Closing the Window

The Preact team responded quickly (and correctly) by reverting the loose check and restoring strict type validation. The patch essentially ensures that VNodes must be created by the framework's internal factory methods, which stamp them with symbols or non-serializable properties that cannot be spoofed via JSON.

Remediation Steps:

  1. Update Immediately: Move to 10.26.10, 10.27.3, or 10.28.2 depending on your branch.
  2. Audit Your Renders: Search your codebase for usage of {variable} where variable comes directly from an API response without validation.
  3. Defensive Coding: Even after patching, avoid passing raw objects to the render tree. Cast them explicitly:
// Do this:
<div>{String(userBio)}</div>
 
// Or this (if you expect an object structure):
<div>{userBio.text}</div>

This is a stark reminder: libraries are code, and code has bugs. Trust, but verify. And maybe don't trust JSON quite so much.

Official Patches

PreactOfficial GitHub Advisory and Patch Details

Technical Appendix

CVSS Score
7.2/ 10
CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:U
EPSS Probability
0.05%
Top 85% most exploited

Affected Systems

Preact Framework (10.26.x, 10.27.x, 10.28.x)Single Page Applications (SPAs) using PreactWeb components built with Preact

Affected Versions Detail

Product
Affected Versions
Fixed Version
Preact
Preact
10.26.5 - 10.26.910.26.10
Preact
Preact
10.27.0 - 10.27.210.27.3
Preact
Preact
10.28.0 - 10.28.110.28.2
AttributeDetail
CWE IDCWE-843 (Type Confusion)
Attack VectorNetwork (JSON Payload)
CVSS v3.16.1 (Medium)
CVSS v4.07.2 (High)
ImpactCross-Site Scripting (XSS)
Exploit StatusPoC Available

MITRE ATT&CK Mapping

T1059.007Command and Scripting Interpreter: JavaScript
Execution
T1190Exploit Public-Facing Application
Initial Access
CWE-843
Access of Resource Using Incompatible Type ('Type Confusion')

The product accesses a resource using an incompatible type, triggering a logic error that allows arbitrary code execution or component injection.

Known Exploits & Detection

Internal ResearchJSON payload demonstrating VNode spoofing via `__v: 0` property.

Vulnerability Timeline

CVE-2026-22028 Published
2026-01-08
GHSA-36hm-qxxp-pg3m Advisory Released
2026-01-08
Patches Issued (10.26.10, 10.27.3, 10.28.2)
2026-01-08

References & Sources

  • [1]Preact Release Notes
  • [2]NIST CVE-2026-22028 Entry

Attack Flow Diagram

Press enter or space to select a node. You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.
Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.

More Reports

•7 minutes ago•CVE-2026-56677
8.6

CVE-2026-56677: Unauthenticated Server-Side Request Forgery in 9Router OIDC Test Endpoint

A high-severity security vulnerability exists in 9Router, an AI router and token saver dashboard. When dashboard authentication features are disabled or left in default configurations, the application exposes administrative testing routines directly to the public internet. Unauthenticated network adversaries can exploit the OIDC configuration validation endpoint to initiate arbitrary HTTP requests, routing unauthorized traffic to local loops, adjacent container ports, and cloud resource metadata interfaces.

Amit Schendel
Amit Schendel
0 views•5 min read
•about 1 hour ago•CVE-2026-64849
9.3

CVE-2026-64849: Server-Side Request Forgery (SSRF) in MLflow Webhooks via DNS Rebinding

CVE-2026-64849 is a critical Server-Side Request Forgery (SSRF) vulnerability affecting MLflow tracking servers prior to version 3.15.0. It allows unauthenticated remote attackers to bypass outbound request destination filters using DNS rebinding or HTTP redirects. This exposure risks compromising sensitive cloud infrastructure metadata and internal microservices.

Alon Barad
Alon Barad
2 views•5 min read
•about 2 hours ago•CVE-2026-69146
6.5

CVE-2026-69146: Missing Authorization Bypass in MLflow Basic Authentication Middleware

This technical report details a missing authorization vulnerability (CVE-2026-69146 / GHSA-3p64-6gvh-82v5) affecting the MLflow platform from version 3.13.0 to 3.15.0. When MLflow is configured with the built-in basic-auth plugin, authenticated users can bypass run-level UPDATE authorization checks, enabling unauthorized dataset and model lineage metadata injection.

Alon Barad
Alon Barad
2 views•7 min read
•about 3 hours ago•CVE-2026-69148
7.1

CVE-2026-69148: Broken Object Level Authorization (BOLA) in MLflow Model Registry

MLflow prior to version 3.15.0 fails to perform proper authorization checks when registering model versions, allowing authenticated users with access to a registered model to link and access artifacts from runs and models belonging to other users without authorization.

Amit Schendel
Amit Schendel
4 views•7 min read
•about 4 hours ago•CVE-2026-59893
7.5

CVE-2026-59893: Regular Expression Denial of Service in sqlparse Lexer

A high-severity Regular Expression Denial of Service (ReDoS) vulnerability in the sqlparse Python library prior to version 0.6.0 allows unauthenticated remote attackers to trigger CPU exhaustion and application denial of service via crafted SQL inputs containing unmatched dollar-quoted literals or unclosed multiline comments.

Alon Barad
Alon Barad
4 views•6 min read
•about 5 hours ago•GHSA-FHGH-WQ4Q-R37X
7.8

GHSA-FHGH-WQ4Q-R37X: Remote Code Execution via Sigstore Signature Verification Bypass in uniget CLI

A high-severity logic inversion flaw in the uniget CLI completely bypasses Sigstore cryptographic signature verification on metadata files by default. If an attacker can poison the package metadata cache or repository, they can execute arbitrary OS commands under the privileges of the active user.

Alon Barad
Alon Barad
5 views•5 min read