CallSy is an AI-powered Android scam call detection app that protects users from fraudulent phone calls by analyzing speech, detecting scam-related keywords, and providing real-time alerts β all with a strong focus on privacy and explainability.
- β Real-Time Scam Call Detection
- β Speech-to-Text using OpenAI Whisper (locally or via API)
- β Scam Keyword & Sentiment Analysis via LLaMA 3.2 NLP
- β Highlights Risky Phrases (e.g., "OTP", "CVV", "bank")
- β Explainable AI Output with Confidence Score
- β Automatic Post-Call Analysis & Flagging
- β Secure, Offline & Self-hosted Backend Integration
- β Scam History Dashboard & Call Block Option
- β Multi-language & Lightweight Android Support
- Frontend: Android
- AI Tools: OpenAI Whisper (Speech-to-Text), LLaMA 3.2 (NLP), Sentiment Analysis
- Backend: Flask
- Clone the repo or download the APK.
- Allow permissions (Microphone, Storage, Caller ID).
- App automatically runs in background and analyzes call after completion.
- View alerts in dashboard.
Clone the project:
git clone https://github.com/scriptbyayush/ScamShield.git
cd ScamShieldAndroid Studio:
- Open the project in Android Studio.
- Sync Gradle.
- Update API keys and endpoint in
Constants.kt. - Connect your device/emulator and run the app.
// After call ends
String audioPath = AudioRecorder.getLastCallAudio();
String transcript = WhisperTranscriber.transcribe(audioPath);
ScamResult result = NLPAnalyzer.analyze(transcript);
if(result.isScam()) {
AlertManager.showScamWarning(result);
}| Sample Output |
|---|
![]() |
# Create virtual environment
python -m venv venv
source venv/bin/activate
# Install dependencies
pip install -r requirements.txt
# Run server
python app.py- OpenAI Whisper β Speech to Text
- LLaMA 3.2 NLP β Contextual Language Understanding
- Custom Sentiment Analyzer β Detect tone and emotional manipulation
- Keyword Pattern Engine β Detects sensitive terms (e.g., OTP, PAN, CVV)
- No third-party server calls unless explicitly configured
- Local device processing available for Whisper + NLP
- User consent taken for microphone access
- Transparent scam detection logic
- π OTP Interception Protection
- π Cloud Sync for History (Encrypted)
- π§ Continuous Learning from User Feedback
- π§ Email: [email protected]
- π Website: www.callsy.ai (Coming Soon)
