Skip to main content

Security

The server is the trust boundary​

Anything in the browser can be bypassed: an attacker can post to your endpoint without using the editor. Sanitise on the server when saving and when outputting — see Saving and displaying safely.

Shadow DOM isolates the editor from your page's CSS and scripts; it is not a security boundary.

Two layers, one allowlist​

LayerWhenHow
BrowserBefore stored content is shown in the editorParsed in an inert document where nothing runs, cleaned, then inserted as nodes
ServerOn save and on outputsymfony/html-sanitizer with PHP 8.4's native HTML5 parser

Both layers read the same allowlist, so what the editor shows is what the server keeps.

Why PHP 8.4​

A sanitiser that reads HTML differently from browsers can be bypassed through the difference (mutation XSS). PHP 8.4's native HTML5 parser builds the same tree browsers do; older parsers were found to disagree on real cases.

How it is tested​

The same set of cases — formatting that must survive, and attack payloads (onerror, javascript: URLs, <svg>/<noscript>/<math> parsing tricks) — runs in Chromium, Firefox and WebKit, and in PHP, and every environment must produce the same result.

Image uploads​

Uploaded files are identified by their content, SVG is refused, sizes and pixel counts are limited before decoding, and every image is re-encoded, which drops metadata and anything hidden in the file. Uploads require a logged-in user and are rate-limited by default. Details: Image uploads.

Content-Security-Policy​

The editor works under a strict policy such as default-src 'self'; style-src 'self'; script-src 'self': it uses no inline scripts and styles its shadow tree with constructable stylesheets, which style-src does not block.

Reporting a vulnerability​

Please do not open a public issue. Use GitHub's private vulnerability reporting on the repository, or email support@2bo-velocity.com with "Security" in the subject.