Turn SVG uploads into plain images and refuse HTML

How to do it

An SVG file is a document, not a picture. It can hold scripts, event handlers, links to other files and external entities. A browser that opens it directly runs all of that. An HTML upload is a whole web page under your name, which also makes it a ready phishing page. Converting to PNG in a sandbox, as Run media tools in a sandbox with no network says, drops every script. If the SVG itself must be kept, clean it with an SVG allowlist sanitiser. Serve it only from the domain in Serve user files from their own domain. Detect the type from the bytes, not the name or the declared type.

Examples

JavaScript example

import sharp from 'sharp';
import { fileTypeFromBuffer } from 'file-type';

const LOGO_WIDTH = 512;
const VIEWABLE_TYPES = new Set(['image/png', 'image/jpeg', 'image/webp']);

export async function acceptLogo(bytes) {
  const detected = await fileTypeFromBuffer(bytes);
  if (detected?.mime === 'application/xml' || bytes.subarray(0, 512).toString().includes('<svg')) {
    return sharp(bytes, { limitInputPixels: 25_000_000 }).resize(LOGO_WIDTH).png().toBuffer();
  }
  if (!VIEWABLE_TYPES.has(detected?.mime)) throw new UnsupportedMediaType(detected?.mime);
  return bytes;
}

Python example

import cairosvg
import magic

LOGO_WIDTH = 512
VIEWABLE_TYPES = {"image/png", "image/jpeg", "image/webp"}

def accept_logo(data: bytes) -> bytes:
    detected = magic.from_buffer(data, mime=True)
    if detected == "image/svg+xml":
        return cairosvg.svg2png(bytestring=data, output_width=LOGO_WIDTH, unsafe=False)
    if detected not in VIEWABLE_TYPES:
        raise UnsupportedMediaType(detected)
    return data

How to check

Upload logo.svg holding <script> and onload= to the test copy. The stored file is a PNG, or an SVG with no script, handler or external reference. Upload page.html renamed to photo.jpg, and it is refused with 415.

Sources

• OWASP File Upload Cheat Sheet
• CWE-79 Improper Neutralization of Input During Web Page Generation

Related rules

• MED-14 Run media tools in a sandbox with no network in Files and media
• MED-3 Serve user files from their own domain in Files and media