@sentinelrn/native

Native runtime-integrity detectors for SentinelRN, with Android (Kotlin) and iOS (Swift) implementations.

This package implements the IntegrityProvider interface from @sentinelrn/core. It collects a flat "integrity snapshot" from the device and maps it into scored threat signals — all severity/confidence/scoring lives in core, keeping detection consistent across platforms.

Install

pnpm add @sentinelrn/core @sentinelrn/native

For bare React Native, install pods:

cd ios && pod install

Usage

Register the provider once at startup, before the first integrity.check():

import { SentinelRN } from "@sentinelrn/core";
import { registerSentinelNative } from "@sentinelrn/native";

registerSentinelNative();

const report = await SentinelRN.integrity.check();

What it detects

| Platform | Signals | | --- | --- | | Android | root (su binaries, root packages, test-keys, writable system paths), emulator, debugger, developer mode, mock location, hooking (Frida/Xposed) | | iOS | jailbreak (filesystem paths, sandbox escape, Cydia scheme), simulator, debugger (P_TRACED) |

Graceful degradation

When the native module is unavailable (Expo Go, web, tests), the provider falls back to best-effort JS-only heuristics and never throws — a detection gap becomes missing signals, not a crash. Meaningful root/jailbreak/hooking detection requires the native module in a development or production build.

Architecture support

Works on both the legacy bridge and the new architecture (Fabric / TurboModules, bridgeless).

• A codegen spec ships at src/NativeSentinelRNNative.ts and is wired via the codegenConfig block in package.json, so react-native-codegen generates typed TurboModule interfaces at app build time.
• At runtime the module is resolved through TurboModuleRegistry.get first, then NativeModules — so the same provider works under either architecture, and returns null (degrading gracefully) when neither is present.

Detection is heuristic and an arms race. SentinelRN raises attacker cost; it does not guarantee detection. See the threat model.

MIT © SentinelRN