OffGuard: AI-Assisted Integrity Monitoring for Digital Examinations

OffGuard is a distributed integrity monitoring engine designed to secure digital examinations in bandwidth-constrained and offline-prone environments. Unlike traditional proctoring solutions that rely on continuous video streaming and cloud processing, OffGuard decentralizes the monitoring layer, processing AI-verified events locally on exam machines.

The Vision

Modern examination infrastructures are vulnerable to central points of failure: internet outages, server overloads, and privacy risks. OffGuard introduces a resilient "Offline-First" monitoring layer that ensures exam integrity persists even when the connection fails.

Tech Stack

Backend (The Coordination Layer)

• Runtime: Node.js (TypeScript)
• Framework: Express
• Database: PostgreSQL
• Infrastructure: Docker & Docker Compose
• Security: Hybrid AES-RSA encryption for payload delivery and log synchronization.

Frontend (The Administration Portal)

• Framework: Next.js (TypeScript)
• Styling: Tailwind CSS
• Desktop Wrapper: Tauri (for secure lockdown browser capabilities and local system access)

Intelligence (The OffCode Engine)

• Language: Python
• Computer Vision: YOLOv8 (Real-time object detection)
• Logic: Local AI modules for:
  • Multi-face detection
  • Mobile phone/Object detection
  • Audio anomaly analysis
  • Active window and process monitoring

System Architecture

OffGuard acts as an intelligent middleware between students and exam platforms (e.g., Pearson VUE, TCS iON).

graph TD
    A["Student<br/>Terminal"]
    B["OffCode<br/>AI Engine"]
    C["Encrypted<br/>Log Store"]
    D["OffGuard<br/>Backend"]
    E["Admin<br/>Dashboard"]

    A -->|Local Monitoring| B
    B -->|AI Verified Logs| C
    C -->|Auto Sync| D
    D -->|Integrity Insights| E

    subgraph "Student Machine"
        B
        C
    end

    subgraph "Central Infrastructure"
        D
        E
    end

Loading

Key Features

1. Offline Resilience

The monitoring engine continues to function during network outages. Integrity logs are generated and stored locally in a tamper-proof encrypted format, synchronizing automatically once connectivity is restored.

2. Privacy-Conscious Monitoring

Instead of full video surveillance, OffGuard records only "suspicious event" snippets and metadata. This significantly reduces personal data storage and protects student privacy while maintaining full auditability.

3. Distributed Bandwidth Optimization

By offloading AI processing to the student machine, OffGuard reduces cloud compute costs and eliminates the need for high-bandwidth video streaming across exam centers.

4. Security Architecture

• Question Protection: Question papers are AES-encrypted; the AES key is RSA-wrapped for specific exam centers.
• Tamper-Proof Logs: Every integrity event is hashed and digitally signed using SHA-256 to ensure non-repudiation.
• Access Control: Enforced via asymmetric key pairs, ensuring only authorized centers can decrypt exam material.

Getting Started

Prerequisites

• Docker & Docker Compose
• Python 3.10+ (for local engine development)
• Node.js 18+

Quick Start (For dev):

1. Clone the repository:

   git clone https://github.com/prawinkumar1506/Offguard-ai-proctoring-engine.git

2. Start the Backend Infrastructure:

   docker-compose up -d

3. Initialize the Local Engine: Navigate to /checker and follow the setup instructions in the local requirements.

4. Run the Local Engine:

   cd checker
   python main.py

5. Access the platform:

   cd frontend
   npm install
   cd src-tauri
   npx tauri dev

---

License

Internal project for secure digital examination infrastructure.