WebView & JS-bridge security — WKWebView, React Native

security · folio

In one line: A WebView runs someone’s JavaScript inside your app. The bridge (WebKit message handlers, React Native postMessage, Android addJavascriptInterface) hands that JavaScript native powers — and it is reachable by every script in every frame of whatever page is loaded. Treat each message as untrusted input: check the origin and frame, allowlist commands, restrict navigation, pass data to JS as values, never as code.

Download PDF Print view LaTeX source

WebView & JS-bridge security — WKWebView, React Native — figure 1

iOS

  • UIWebView (deprecated iOS 12) ran web content in your process; the App Store refused new apps using it from Apr 2020. WKWebView: separate WebContent process, Safari’s engine + JIT.
  • Bridge: userContentController.add(handler, name: "app") → JS window.webkit.messageHandlers.app.postMessage(obj) (JSON-like types). The handler gets a WKScriptMessage: body, frameInfo (isMainFrame, request, securityOrigin), world. Reply: WKScriptMessageHandlerWithReply (iOS 14) — JS gets a Promise.
  • Content worlds (iOS 14): .page, .defaultClient, .world(name:); add(_:contentWorld:name:) exposes a handler only in that world.
  • Native → JS: evaluateJavaScript("show('\(name)')") is injection (a quote in name ends the string). callAsyncJavaScript(_:arguments:in:contentWorld:) passes values.
  • Navigation: decidePolicyFor — allow https + your hosts; others → Safari; cancel javascript:, file:, data:, custom schemes. iOS 14 WKAppBoundDomains + limitsNavigationsToAppBoundDomains disable script injection / handlers elsewhere.
  • Files: loadFileURL(_:allowingReadAccessTo:) — the narrowest directory, never the whole container. Debug: isInspectable (iOS 16.4, default false) — keep it off in release.
  • The controller retains its handlers → handler → VC → WebView cycle: weak proxy, or removeScriptMessageHandler(forName:).

Android

addJavascriptInterface: every @JavascriptInterface method, to every frame and origin (before API 17: reflection → RCE). Prefer AndroidX WebViewCompat.addWebMessageListener with allowedOriginRules — the listener gets sourceOrigin + isMainFrame. setAllowFileAccess defaults false from API 30. shouldOverrideUrlLoading = decidePolicyFor.

Example — a checked bridge (Swift) and RN

final class Bridge: NSObject, WKScriptMessageHandlerWithReply {
  func userContentController(_ c: WKUserContentController,
      didReceive m: WKScriptMessage) async -> (Any?, String?) {
    let o = m.frameInfo.securityOrigin
    guard m.frameInfo.isMainFrame, o.protocol == "https",
          o.host == "app.example.com",
          let cmd = Command(json: m.body)    // typed, allowlisted
    else { return (nil, "denied") }
    return (await perform(cmd), nil)
  }
}
cfg.userContentController.addScriptMessageHandler(Bridge(),
    contentWorld: .page, name: "app")
try await web.callAsyncJavaScript("render(item)",
    arguments: ["item": title], contentWorld: .page)
<WebView source={{ uri: 'https://app.example.com' }}
  originWhitelist={['https://app.example.com']} // navigation only
  onShouldStartLoadWithRequest={r =>
    new URL(r.url).origin === 'https://app.example.com'}
  onMessage={e => { const m = parseCommand(e.nativeEvent.data)
                    if (m && allowed(m)) run(m) }}   // string in
  javaScriptEnabled setSupportMultipleWindows /* keep true */ />

React Native WebView — the traps

  • onMessage receives window.ReactNativeWebView.postMessage(string) — a string only, with no verified sender: any script the page runs can post.
  • originWhitelist (default http://*, https://*) lists origins the WebView may navigate to; others are handed to the OS. It does not filter messages.
  • onShouldStartLoadWithRequest: false blocks; not called on Android’s first load.
  • injectedJavaScriptObject is visible to all frames; injectedJavaScriptForMainFrameOnly defaults true.
  • GHSA-36j3-xxf7-4pqg / CVE-2020-6506 (Sep 2020, moderate): Android WebView before 83.0.4103.106 let a cross-origin iframe run code in the top document (UXSS); react-native-webview ≤ 10.10.2 exposed, 11.0.0 mitigates — setSupportMultipleWindows=false re-opens it.

Traps

  • “It’s our own page” — until a CDN script, an ad or one XSS runs on it.
  • Login / OAuth in an embedded WebView: the app can read the password; RFC 8252 says use the system browser (ASWebAuthenticationSession).
  • Injecting the refresh token into JS for the web part — any page script can read it.

Remember

“Who sent it, may they, is it well-formed?” — then values out, never code.

Report a problem ·Discuss this folio