A cross-platform mobile application framework for phishing detection research. Built with Kotlin Multiplatform, MsgVerify provides an extensible platform for studying cybersecurity awareness on mobile devices.

MsgVerify is a Kotlin Multiplatform (KMP) application framework with a shared codebase deployed on Android and iOS. It integrates the ContextGuard library for on-device machine learning inference and provides extensible architecture for researchers to customise datasets, visual indicators, and detection heuristics.

MsgVerify is designed as an extensible research platform for mobile security studies. Key extension points include:

• Configurable Risk Thresholds — Adjust sensitivity via MsgVerifyConfig.kt
• Swappable Demo Datasets — Replace JSON files with custom threat scenarios
• Custom Heuristics — Register detection rules via HeuristicRegistry
• Themeable UI — Modify visual indicators for A/B testing warning designs
• Cross-Platform Consistency — Identical detection logic on Android and iOS

See EXTENDING.md for detailed documentation on framework customisation.

MsgVerify is built on top of the ContextGuard library, which provides the core content verification engine. ContextGuard is a Kotlin Multiplatform library that encapsulates:

• Machine learning models (PyTorch Lite) for text classification
• URL extraction and analysis algorithms
• Heuristic-based detection rules
• Risk scoring and reasoning generation

MsgVerify demonstrates how to integrate ContextGuard into a real-world application by providing:

• A user-friendly UI for content verification
• Demo scenarios with mock phishing/spam data
• iOS Shortcuts integration for system-level access
• History tracking and result visualization

Key Features:

• Offline Content Verification: Analyzes text for phishing/scam indicators using embedded ML models
• URL Detection & Analysis: Extracts and scores URLs within content
• Multi-Channel Support: SMS, Email, Social Media, Generic Text
• Demo Scenarios: Interactive demonstrations with mock data
• Verification History: Maintains local history of checked items
• iOS Shortcuts Integration: Siri/Shortcuts support for voice commands

• Frontend: Kotlin Multiplatform with Compose Multiplatform
• Architecture Pattern: MVVM + Repository Pattern
• Dependency Injection: Koin
• ML Backend: ContextGuard library (embedded models)
• Platforms: Android (API 30-35), iOS (x64, arm64, simulator)

MsgVerify/
├── composeApp/              # Main Kotlin Multiplatform application
│   ├── src/commonMain/      # Shared code (Android + iOS)
│   ├── src/androidMain/     # Android-specific code
│   ├── src/iosMain/         # iOS-specific code
│   └── src/commonTest/      # Shared tests
├── iosApp/                  # iOS-specific implementation
└── mocks/                   # Test data for demos

User Input
    ↓
HomeViewModel.onVerifyClicked()
    ↓
MsgVerifyRepository.verifyContent()
    ↓
ContentVerifierImpl (from ContextGuard library)
    ├─ ML Model Classification
    │  └─ TextClassificationResult.Safe/Unsafe
    ├─ URL Detection & Scoring
    │  └─ List<Float> urlScores
    └─ Reason Analysis
       └─ List<Reason> reasons
    ↓
LinkVerificationState Updates
    ├─ UI Re-composition
    └─ History Management

Risk thresholds are configurable via MsgVerifyConfig.kt. Default values:

• Green: <40% malicious confidence (Safe)
• Yellow: 40-70% malicious confidence (Medium Risk)
• Red: >70% malicious confidence (High Risk)

See EXTENDING.md for customisation instructions.

• JDK: Java 11 or higher
• Android Studio: Latest stable version with Kotlin plugin
• Xcode: For iOS development (macOS only)

./gradlew :composeApp:assembleDebug

1. Open iosApp/iosApp.xcodeproj in Xcode
2. Select your target device/simulator
3. Build and run (⌘+R)

All dependencies are managed via Gradle:

./gradlew build

Navigate to ios app with cd iosApp and install the dependency

pod install

Create file named Podfile in the iOS app directory. Copy and paste the following:

platform :ios, '13.0'
use_frameworks!

target 'iosApp' do
  pod 'TensorFlowLiteObjC', '2.17.0'
  pod 'TensorFlowLiteObjC/Metal', '2.17.0'
end

Install with pod install

• Compose Multiplatform: 1.8.2
• Material3: 1.8.2
• Lifecycle ViewModel: 2.9.1

• Koin: 4.0.3

• ContextGuard: 1.0.0 (Custom ML-based verification library)

• PyTorch Lite Multiplatform: 0.7.0

The project includes mock data for testing and demonstrations:

mocks/
├── safe/
│   ├── sms.json
│   ├── emails.json
│   └── social_media_messages.json
└── malicious/
    ├── sms.json
    ├── emails.json
    └── social_media_messages.json

MsgVerify supports iOS Shortcuts via VerifyMessageIntent.swift:

1. Add Shortcut: Use the Shortcuts app to create a custom action
2. Voice Command: Say "Hey Siri, verify this message"
3. System Integration: Seamlessly integrates with iOS share sheets

To test with a local build of the ContextGuard library:

1. Edit settings.gradle.kts
2. Set testLocally = true
3. Rebuild the project

• Purpose: Main content verification interface
• Input: Text area for messages/URLs
• Output: Risk assessment with color-coded results
• History: Recent verification history

• SMS Demo: Phishing SMS detection examples
• Email Demo: Phishing email detection examples
• Social Media Demo: Social media content analysis examples