🛡️ SecureVision AI

Privacy-Preserving Federated Cyber Threat Detection Platform

SecureVision AI is a working cybersecurity prototype that demonstrates how AI-based threat detection can be performed without sharing raw endpoint telemetry, using federated learning concepts and privacy-by-design architecture.

Built for the AI + Cybersecurity & Privacy track.

---

🚀 Problem Statement

Traditional cybersecurity solutions:

• Centralize raw endpoint logs and telemetry

• Increase privacy, compliance, and breach risks

• Create a single point of failure

Modern organizations need:

• Real-time threat detection

• Strong privacy guarantees

• Scalable and secure architectures

---

💡 Solution Overview

SecureVision AI solves this by:

• Running AI anomaly detection locally on each endpoint

• Sharing only privacy-safe metadata with the SOC

• Never transmitting raw logs, files, or personal data

This ensures:

• 🔐 Privacy preservation

• ⚡ Real-time detection

• 📊 Centralized SOC visibility

---

🤖 AI & Security Design

Endpoint AI

• Uses Isolation Forest (unsupervised ML)

• Detects unknown and zero-day anomalies

• Lightweight and endpoint-friendly

Privacy-Preserving Design

• No raw telemetry leaves the endpoint

• Only metadata (hostname, features, severity)

• GDPR-friendly and enterprise-ready

Federated Learning Concept

• Local model training on each endpoint

• Central intelligence aggregation

• Privacy preserved by default

---

🖥️ SOC Dashboard Features

• Real-time endpoint activity

• Blinking anomaly alerts

• Severity-based color coding

  • LOW → Green

  • MEDIUM → Yellow

  • HIGH → Red

• Global endpoint visualization (simulated map)

• Auto-refresh every 5 seconds

• Dark / Light mode toggle

• Fully responsive (mobile, tablet, desktop)

---

📸 Screenshots

🖥️ SOC Dashboard — Dark Mode

---

🚨 Endpoint Alert with Hostname

---

📱 Responsive Mobile View

---

🔐 AMD Technology Alignment

SecureVision AI is designed to be:

• Confidential Computing ready

• Compatible with AMD Secure Encrypted Virtualization (SEV)

• Secure for deployment in untrusted cloud environments

---

🧪 How to Run the Project

1. Clone the Github Repository

   git clone https://github.com/kavya-seth-vns/SecureVision-AI.git
   

2. Create Virtual Environment

   python -m venv venv
   
   cd venv\Scripts\
   
   ./Activate
   

3. Install Dependencies

   python -m pip install pandas scikit-learn flask requests
   

4. Start SOC Aggregator

   cd aggregator
   
   python app.py
   

5. Run Endpoint Agent

   cd endpoint
   
   python agent.py
   

6. Open Dashboard

   http://127.0.0.1.5000

   ---

📁 Project Structure

securevision-ai/

├── aggregator/ # Central SOC dashboard & API

├── endpoint/ # Endpoint-side AI agent

├── README.md

└── .gitignore

---

🎯 Why SecureVision AI Stands Out

• Fully working AI prototype (not a mock UI)

• Strong focus on privacy & security

• Real-world SOC use case

• Modern, responsive dashboard

• AMD-aligned confidential computing design

• Scalable and enterprise-ready

---

🚀 Future Enhancements

• True federated model aggregation

• Differential privacy noise injection

• SOC alert acknowledgment workflow

• Role-based access control

• Cloud deployment on confidential VMs

---

🏁 Conclusion

SecureVision AI demonstrates that powerful cybersecurity intelligence can be achieved without compromising user privacy, using federated AI principles and confidential computing concepts.

---

👥 Team

Team Name: SecureVision

Team Member: Kavya Seth , Sristi Seth , Prashant Kumar Srivastava

Hackathon: AMD Slingshot / Hack2Skill

---

AI-powered cybersecurity intelligence without compromising privacy.