Athena Security Guardian

An Edge AI-Powered Virtual Security Assistant Built for the Snapdragon Multiverse Hackathon @ Princeton.

⸻

🚀 Overview

Athena Security Guardian is an AI-powered edge application that redefines physical security for offices, hospitals, residential complexes, and public spaces.

Unlike traditional systems that rely on costly human monitoring or cloud-based AI (which introduce latency and privacy risks), Athena operates entirely on-device using Snapdragon-powered laptops and phones.

This ensures: • ⚡ Ultra-low latency decisions (<50ms) • 🔒 Privacy-first operations (no cloud dependency, no biometric data leaving the device) • 🛡️ Always-on reliability (works even offline)

Athena acts as both: 1. A virtual security guard – monitoring, detecting, and alerting in real-time. 2. An intelligent assistant – answering visitor queries and escalating when needed.

⸻

🧩 Key Features • 👁️ Edge-based Motion & Person Detection • YOLOv8 (quantized ONNX/TFLite optimized for Snapdragon NPU) • Runs at 30+ FPS with <50ms latency • 🧑‍🤝‍🧑 Facial Recognition & Role Classification • MobileFaceNet embeddings for resident/staff vs. unknown visitor differentiation • GDPR/CCPA compliant (no cloud storage of biometrics) • 🎙️ Voice Interaction • On-device Whisper-small (quantized) for speech-to-text • DistilBERT intent classifier for query understanding • On-device TTS for natural responses • 🚨 Security Threat Detection • Identifies tailgating, loitering, restricted entry attempts • Triggers instant audio alerts + optional push notifications • 🛡️ Offline & Privacy-first • Works without internet • Session-only embeddings (no permanent storage)

⸻

🏗️ Technical Architecture

Hardware: Snapdragon-powered Laptop + Mobile Device Vision Pipeline: Camera → Preprocessing → YOLOv8n → Face Embedding → Event Trigger

Audio Pipeline: Mic → Whisper-small → Intent Classifier → Response Generator → On-device TTS

Automation Layer: Handles access control, alerting, and visitor interaction logic.

⸻

🎯 Example Scenarios • Hospitals & Offices – Automates visitor check-in, reduces load on security staff. • Residential Complexes – 24/7 AI gatekeeper, instantly alerts residents. • Events & Public Spaces – Provides multilingual guidance and detects crowd anomalies.

⸻

⚡ Getting Started

🔹 Requirements • Snapdragon-powered Copilot+ PC or Galaxy S25 (hackathon kit) • Python 3.10+ • Dependencies:

pip install -r requirements.txt

🔹 Running Athena 1. Clone this repo:

git clone https://github.com/ankithn30/AthenaSecurityGuardian.git cd AthenaSecurityGuardian

2.	Start the security guardian:

python main.py

3.	Access the local web dashboard at:

http://localhost:5000

⸻

🧪 Demo 1. Point your phone camera (via QR connection) as the external video feed. 2. Ask Athena questions like: • “Where is Room 203?” • “Is this person allowed entry?” 3. Simulate threats (tailgating, loitering) and watch Athena respond in real-time.

⸻

🌍 Impact

Athena demonstrates the power of Snapdragon Edge AI to: • Reduce costs by replacing manual monitoring • Enhance safety and reliability without sacrificing privacy • Scale security for residential, corporate, and public spaces • Even potentially save lives in critical scenarios (e.g., hospitals, emergencies)

⸻

👥 Team

Athena Security Guardian was developed at the Snapdragon Multiverse Hackathon (Princeton, 2025) by: - Satvika Maram -satvika.maram@gmail.com * Ankith Nagabandi - ankithnagabandi@gmail.com

⸻ 📜 License

This project is licensed under the MIT License.