Architecture and security
The application is structured into backend and frontend components, with performance-critical tasks handled in Rust and C++ for speed and safety, and high-level logic in Python and TypeScript for flexibility.
Backend
Python: Orchestrates application logic, Matrix protocol integration, and AI model execution (e.g., Whisper for translation). Uses
pyo3
andctypes
for Rust and C++ bindings.Rust: Handles encryption (Signal Protocol, CRYSTALS-Kyber, AES-256-GCM), WebRTC peer connections, and session management with Tor for anonymous sessions.
C++: Implements low-level audio processing (noise reduction, echo suppression, voice correction) and deepfake detection using RNNoise and OpenCV.
Components:
Core: Manages sessions, WebRTC, and configuration.
Encryption: Implements hybrid encryption and watermarking.
AI: Processes audio and video with on-device AI models.
Security: Handles biometric authentication and SOS broadcasts.
Frontend
TypeScript/React: Provides a responsive UI with components for calls, chats, settings, and SOS features.
WebRTC Hooks: Manage low-latency audio/video streams.
Tailwind CSS: Ensures a modern, consistent design.
Security Measures
Selenium AI Call prioritizes security through the following mechanisms:
Local Rendering: All AI processing (audio, translation, deepfake detection) occurs on-device, minimizing data exposure to servers.
Anonymous Sessions (Ghost Mode): Uses one-time identifiers and Tor-based routing to prevent tracking.
Hybrid Encryption:
Signal Protocol for E2E encryption with Perfect Forward Secrecy.
CRYSTALS-Kyber for post-quantum security.
ZRTP for dynamic key sharing during calls.
Attachment Encryption: Automatically encrypts photos and documents using AES-256-GCM.
Self-Destructing Messages: Messages are deleted if a screenshot is attempted, protecting sensitive content.
Biometric Authentication: Voice and facial analysis verify user identity at login.
Digital Watermarks: Ensure audio/video authenticity, preventing tampering.
Scalability and Reliability
WebRTC Modifications: Achieve <100ms latency for real-time communication.
Matrix Protocol: Enables interoperability with other messengers (e.g., Signal, Telegram).
Modular Design: Facilitates updates and community contributions.
Last updated