https://github.com/mensonones/expo-ai-context-shield
The ultimate privacy layer for Expo apps. Blocks AI OCR, screen captures, and system context leakage with zero performance overhead.
https://github.com/mensonones/expo-ai-context-shield
ai expo react-native security
Last synced: 5 months ago
JSON representation
The ultimate privacy layer for Expo apps. Blocks AI OCR, screen captures, and system context leakage with zero performance overhead.
- Host: GitHub
- URL: https://github.com/mensonones/expo-ai-context-shield
- Owner: mensonones
- License: mit
- Created: 2025-12-31T16:07:13.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2026-01-01T12:22:04.000Z (6 months ago)
- Last Synced: 2026-01-06T09:35:03.193Z (6 months ago)
- Topics: ai, expo, react-native, security
- Language: Kotlin
- Homepage:
- Size: 193 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# 🛡️ Expo AI Context Shield
**The ultimate privacy layer for Expo apps. Protect sensitive UI from AI vision, Screen Captures (iOS), and System Intelligence.**
In 2025/2026, OS-level AI (like Google’s Gemini Nano and Samsung’s Galaxy AI) can "read" your app's screen to provide context. While helpful, this poses a massive privacy risk for apps handling bank balances, health data, or private credentials.
`expo-ai-context-shield` provides a **semantic security boundary** that makes your content invisible to AI OCR and system intelligence without compromising user experience or causing rendering bugs.
> [!WARNING]
> **iOS Support:** The iOS implementation is based on native security APIs but has **not been tested on physical devices yet**. Contributions and feedback are welcome!
---
## 🚀 Key Features
* **AI Vision Blocking:** Prevents on-device AI (Circle to Search, Gemini Nano, Galaxy AI) from indexing or "reading" your sensitive UI.
* **Hardware-Level Protection:** Automatically blocks screenshots and screen recordings on both iOS and Android using native security layers.
* **Zero Rendering Bugs:** Optimized for React Native to ensure your content never disappears or "blinks" when security is toggled.
* **Zero Performance Overhead:** Pure native implementation with no background listeners or heavy wrappers.
* **Ultra Lightweight:** Only ~9kB (gzipped), ensuring no impact on your app's bundle size.
---
## 📦 Installation
```bash
npx expo install expo-ai-context-shield
```
> [!IMPORTANT]
> **This is a native module.** It requires [Development Builds](https://docs.expo.dev/development/introduction/) and will not work in Expo Go. You must run `npx expo prebuild` to use it in your project.
---
## 💡 Why this module?
### The Privacy Risk in the Gen-AI Era
You might think: *"If the user triggered the AI, they know what they are doing"*. In reality, the risk is more subtle:
* **Context Leakage:** When a user uses *Circle to Search* to identify a shoe, the AI captures the **entire screen**. If a bank balance or CPF is visible next to it, that sensitive data is sent to AI servers as "context".
* **Activity Logs:** AI assistant captures are often saved in the user's account history (Google/Samsung/Apple). Without protection, your app's private data ends up stored permanently in third-party logs.
* **Always-on Snapshots:** Features like "Recall" or "System Timelines" take periodic screenshots of the screen. This module ensures your app is a "black hole" in those automated captures.
### Comparison
| Feature | Standard View | `FLAG_SECURE` (Full App) | **AI Context Shield** |
| --- | --- | --- | --- |
| Protects against AI | ❌ | ✅ | ✅ |
| Blocks Screenshots | ❌ | ✅ | ✅ |
| UX Friendly | ✅ | ❌ | ✅ |
| OCR Resistant | ❌ | ✅ | ✅ |
---
## ⚡ Performance & Size
This module is designed to be as transparent as possible:
- **Package Size:** ~8.6 kB (compressed) / ~24 kB (unpacked).
- **Native Footprint:** Uses standard system APIs (`FLAG_SECURE` on Android, `isSecureTextEntry` on iOS).
- **No JS Bridge Overhead:** Communication is minimal and only happens when the `isSensitive` prop changes.
---
## 🛠 Usage
Simply wrap your sensitive components with the `AiContextShield`.
```tsx
import { AiContextShield } from 'expo-ai-context-shield';
import { Text, View } from 'react-native';
export default function SensitiveScreen() {
return (
Public information (visible to AI)
{/* This part will be blacked out in screenshots and invisible to Gemini/Galaxy AI */}
Balance: $1,250,000.00
);
}
```
> [!TIP]
> You can use `` as a shorthand for ``. In JSX, if you pass a prop without a value, it defaults to `true`.
>
> **Why does this work without children?**
> On Android, this module uses `FLAG_SECURE` on the **entire Activity Window**. This means that simply mounting a single `` anywhere in your component tree will protect the **entire screen** from screenshots and AI analysis, even if it wraps nothing.
### Screen-Level Protection (No Children)
You can use the component without children to toggle protection for the entire screen (Android).
```tsx
import { AiContextShield } from 'expo-ai-context-shield';
import { View, Text } from 'react-native';
export default function SecretScreen() {
return (
{/* This will trigger FLAG_SECURE on Android, protecting the whole screen */}
This entire screen is now protected.
);
}
```
---
## 🏥 Common Use Cases
| Industry | What to protect? |
| --- | --- |
| **Fintech** | Account balances, Credit Card numbers, CVV, QR Codes. |
| **Healthcare** | Patient names, diagnoses, exam results. |
| **Messaging** | Private chat content, self-destructing photos. |
| **Auth** | Recovery keys, 2FA codes, temporary passwords. |
---
## ⚙️ API Reference
### `AiContextShield` (Component)
| Prop | Type | Default | Description |
| --- | --- | --- | --- |
| **`isSensitive`** | `boolean` | `false` | When `true`, enables hardware-level protection. |
| **`style`** | `ViewStyle` | `undefined` | Standard React Native styles. |
| **`children`** | `ReactNode` | `required` | The content you want to shield. |
---
## 📱 Platform Implementation
* **Android (API 29+):** Uses `IMPORTANT_FOR_CONTENT_CAPTURE_NO_EXCLUDE_DESCENDANTS` to block data indexing and `FLAG_SECURE` to block visual pixel capture. Tested on Samsung S23/S24 (Galaxy AI) and Pixel 8/9 (Gemini).
* **iOS (13.4+):** Implements a hidden `UITextField` canvas layer trick. *(Note: Not yet tested on physical devices)*.
* **Web:** Transparent fallback. Renders a standard `View` to maintain layout consistency.
---
## 🤝 Contributing
We are looking for help with:
* Improving the instance counter for `FLAG_SECURE` on Android.
* iOS screenshot detection events.
1. Fork the repo.
2. Install dependencies: `npm install`.
3. Test in the example app: `cd example && npx expo run:android`.
---
## 📄 License
MIT © @mensonones