feat(security): confirm before opening a link whose text names a different site (#122)

utils/linkSafety.ts compares the registrable domain the visible text claims
(when it looks like a URL/host) with the href's; a mismatch, or a punycode
(IDN) destination, renders the anchor as SuspiciousLink, whose click opens a
confirm — "It shows matrix.lotusguild.org but goes to evil.example." with the
full URL, Cancel / Open anyway (opens in a new tab with noopener). Honest links
are untouched: same registrable domain (youtube.com text over www.youtube.com,
bbc.co.uk over news.bbc.co.uk), plain-word text, mailto:, matrix.to and Lotus
permalinks, anchors with non-text children. Comparator unit-tested (incl. a
Cyrillic-а paypal homograph); verified headless that the phish and IDN
messages are flagged, the honest ones are not, the click shows the confirm and
does not navigate, Cancel keeps you put, Open anyway opens the real target.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPmy3tPq869XDW4njjVaKA

src/app/utils/linkSafety.test.ts

@@ -0,0 +1,40 @@
+import { test } from 'node:test';
+import assert from 'node:assert/strict';
+import { analyzeLink, hostFromText, registrableDomain } from './linkSafety';
+
+test('visible text that names another site is a mismatch', () => {
+  const r = analyzeLink('https://matrix.lotusguild.org/login', 'https://evil.example/login');
+  assert.equal(r?.mismatch, true);
+  assert.equal(r?.shownHost, 'matrix.lotusguild.org');
+  assert.equal(r?.realHost, 'evil.example');
+});
+
+test('honest links are not flagged: same registrable domain, plain words, non-http', () => {
+  assert.equal(
+    analyzeLink('youtube.com/watch?v=1', 'https://www.youtube.com/watch?v=1')?.mismatch,
+    false,
+  );
+  assert.equal(analyzeLink('bbc.co.uk', 'https://news.bbc.co.uk/x')?.mismatch, false);
+  assert.equal(analyzeLink('click here', 'https://evil.example')?.mismatch, false);
+  assert.equal(analyzeLink('evil.example', 'mailto:someone@evil.example'), null);
+  assert.equal(analyzeLink('elsewhere.org', 'https://matrix.to/#/#room:x'), null);
+  assert.equal(
+    analyzeLink('lotusguild.org', 'https://chat.lotusguild.org/home/!r:x')?.mismatch,
+    false,
+  );
+});
+
+test('registrable domain handles two-label suffixes', () => {
+  assert.equal(registrableDomain('news.bbc.co.uk'), 'bbc.co.uk');
+  assert.equal(registrableDomain('www.example.com'), 'example.com');
+  assert.equal(registrableDomain('example.com'), 'example.com');
+});
+
+test('hostFromText only accepts URL/host-shaped text; IDN becomes punycode', () => {
+  assert.equal(hostFromText('Visit https://a.example/path'), null);
+  assert.equal(hostFromText('a.example/path'), 'a.example');
+  assert.equal(hostFromText('user@a.example:8448/x'), 'a.example');
+  assert.equal(hostFromText('pаypal.com'), 'xn--pypal-4ve.com'); // Cyrillic а
+  assert.equal(analyzeLink('paypal.com', 'https://xn--pypal-4ve.com/')?.punycode, true);
+  assert.equal(analyzeLink('paypal.com', 'https://xn--pypal-4ve.com/')?.mismatch, true);
+});

src/app/utils/linkSafety.ts

@@ -0,0 +1,99 @@
+/**
+ * [Gitea #122] Phishing-shape detection for rendered links: the visible text
+ * looks like a URL/hostname but the real destination is a different site.
+ *
+ *   <a href="https://evil.example/login">https://matrix.lotusguild.org/login</a>
+ *
+ * Compares registrable domains (so `youtube.com/x` text over a `www.youtube.com`
+ * href is honest), ignores non-http(s) targets, and surfaces punycode hosts
+ * so an IDN homograph can't hide behind a familiar-looking label.
+ */
+
+// Two-label public suffixes we care about; everything else is eTLD+1 = last 2 labels.
+const TWO_LABEL_SUFFIXES = new Set([
+  'co.uk',
+  'org.uk',
+  'ac.uk',
+  'gov.uk',
+  'me.uk',
+  'ltd.uk',
+  'plc.uk',
+  'net.uk',
+  'com.au',
+  'net.au',
+  'org.au',
+  'edu.au',
+  'gov.au',
+  'co.nz',
+  'org.nz',
+  'net.nz',
+  'co.jp',
+  'ne.jp',
+  'or.jp',
+  'ac.jp',
+  'com.br',
+  'net.br',
+  'org.br',
+  'co.in',
+  'net.in',
+  'org.in',
+  'co.za',
+  'org.za',
+  'com.mx',
+  'com.ar',
+  'com.tr',
+  'com.cn',
+  'com.hk',
+  'com.sg',
+  'com.tw',
+]);
+
+export const registrableDomain = (host: string): string => {
+  const labels = host.toLowerCase().replace(/\.$/, '').split('.');
+  if (labels.length <= 2) return labels.join('.');
+  const lastTwo = labels.slice(-2).join('.');
+  return TWO_LABEL_SUFFIXES.has(lastTwo) ? labels.slice(-3).join('.') : lastTwo;
+};
+
+const HOST_LIKE =
+  /^(?:[a-z][a-z0-9+.-]*:\/\/)?(?:[^\s/?#@]+@)?([a-z0-9¡-￿-]+(?:\.[a-z0-9¡-￿-]+)+)(?::\d+)?(?:[/?#]|$)/i;
+
+/** The hostname a piece of visible text CLAIMS to point at, if it looks like one. */
+export const hostFromText = (text: string): string | null => {
+  const t = text.trim();
+  if (!t || /\s/.test(t)) return null;
+  const m = HOST_LIKE.exec(t);
+  if (!m) return null;
+  try {
+    // Normalise through URL so unicode hosts become punycode like real hrefs do.
+    return new URL(`http://${m[1]}`).hostname;
+  } catch {
+    return null;
+  }
+};
+
+export type LinkSafety = {
+  /** The visible text names a different site than the destination. */
+  mismatch: boolean;
+  shownHost?: string;
+  realHost: string;
+  /** Destination host contains an IDN (punycode) label. */
+  punycode: boolean;
+};
+
+export const analyzeLink = (text: string, href: string): LinkSafety | null => {
+  let url: URL;
+  try {
+    url = new URL(href);
+  } catch {
+    return null;
+  }
+  if (url.protocol !== 'http:' && url.protocol !== 'https:') return null;
+  if (url.hostname === 'matrix.to') return null;
+  const realHost = url.hostname;
+  const punycode = realHost.split('.').some((l) => l.startsWith('xn--'));
+  const shownHost = hostFromText(text) ?? undefined;
+  const mismatch =
+    shownHost !== undefined && registrableDomain(shownHost) !== registrableDomain(realHost);
+  return { mismatch, shownHost, realHost, punycode };
+};