AI Output Rendered in WebView (XSS / local-file read)

← AI + Mobile (LLM x IPC / WebView)

ai_output_to_webview_xss   HARD

Description

LLM output is injected into a WebView via loadHtmlString/evaluateJavascript with no encoding, so model-produced (attacker-influenced) HTML/JS executes in the app origin (AI-output->XSS class; FAQ-Bot CVE-2025-63639 / ZOLL ePCR CVE-2025-12699).

How it works

The assistant output, which an attacker steered via the prompt, is injected into a WebView through loadHtmlString with no output encoding. Because the model text can carry a script tag, that markup executes in the app WebView origin, where it can reach JS bridges or read local files (the AI-output-to-XSS class; FAQ-Bot CVE-2025-63639 / ZOLL ePCR CVE-2025-12699). The demo drives a model and loads its unescaped output into a real system WebView, then reads document.title back out to prove the injected script ran. The secure path HTML-escapes the model output first, so the markup is inert text and nothing executes.

How to exercise it. DVMA is the harness - open this module from the home index and tap the demo action. The screen ships the malicious input and simulates the attacker (e.g. the companion app, crafted intent, or scanned payload) in-process, and the evidence panel prints the proof. The Tools (optional) and Attack inputs below are only needed to reproduce the exploit end-to-end on a real device.

Exploit steps

1. Set up. Build DVMA with a flavor that enables the AI + Mobile (LLM x IPC / WebView) category (e.g. --dart-define-from-file=config/flavors/dev.json) and run on an emulator/simulator you control. The demo needs no external tooling; for the optional on-device reproduction the relevant tools are: garak, promptfoo, frida, objection.
2. Locate the target. From the home index, open AI Output Rendered in WebView (XSS / local-file read) (ai_output_to_webview_xss). The How it works section above describes this module’s specific weakness; the screen states the intended-secure behavior and exposes the vulnerable action.
3. Exploit. Deliver attacker content across the mobile boundary (deep link / clipboard / QR) into the assistant, or steer the model’s OUTPUT into a WebView / intent / tool call, and confirm it executes with no validation boundary in between.
4. Observe the evidence. Trigger the vulnerable action and read the evidence panel - it prints the concrete proof (leaked value, accepted replay, executed payload, or unauthorized result).
5. Contrast with the secure path. Run the module’s secure/hardened action (where provided) and confirm the same attack is rejected - this is what a correct implementation should do.
6. Map it back. Cross-reference the MASTG v2 test(s) MASTG-TEST-0031, MASTG-TEST-0033 for the canonical procedure and remediation.

Tools (optional)

• garak
• promptfoo
• frida
• objection

garak / promptfoo / MCP scanner test the LLM or MCP endpoint behind the app, not the app binary. Point them at the backend model API (find it with mitmproxy / Burp Suite) or a local on-device model server; the in-app demo already exercises the same prompt path.

Attack inputs

Payloads/artifacts you author for the on-device attack. The demo already ships and simulates these in-process (e.g. the malicious companion app / crafted intent is emulated inside the screen), so you only need to craft them to reproduce the exploit on a real device:

• crafted prompt

Real-world references

Concrete public disclosures that match this vulnerability class:

• CVE-2025-63639 (FAQ Bot AI chat XSS)
• CVE-2025-12699 (ZOLL ePCR iOS WebView local-file read)