Skip to content

Vanguard is an Office 365 Add-in for advanced email scanning to prevent malicous attacks on an enterprise scale

Notifications You must be signed in to change notification settings

Nottlespike/Vanguard_L3_70B

Repository files navigation

Project Summary: Vanguard - AI-Powered Email Security Add-in For Data Sensitive Enterprises

  1. Project Overview: We have developed "Vanguard," an innovative Outlook add-in that leverages AI to enhance email security. This tool is designed to analyze emails in real-time for potential security threats, providing users with immediate feedback on the safety of their incoming messages.

  2. Key Features:

    • Real-time email analysis for security threats
    • Integration with an OpenAI-compatible API's in this case Llama 3 70b trained on a small dataset of sythetic for advanced threat detection
    • High level secure interface within Outlook
    • Dynamic light/dark mode support for better accessibility
    • Detailed explanations of security assessments
  3. Technologies Used:

    • JavaScript for core functionality
    • HTML and CSS for the user interface
    • Office.js API for Outlook integration for enterprise use cases
    • Axios for making API calls
    • Webpack for bundling and build processes
    • Node.js for the development environment
  4. Development Achievements:

    • Successfully set up an Office Add-in development environment 100% local API with zero internet access
    • Implemented a taskpane interface in Outlook
    • Integrated with an AI service (OpenAI-compatible endpoint) for email analysis at a complex level
    • Developed a responsive UI that adapts to Outlook's theme settings
    • Implemented error handling and input sanitization for security
    • Created a build process using Webpack for production-ready code
  5. User Experience:

    • Simple one-click analysis of emails
    • Clear visual indicators for email safety status
    • Detailed explanations of AI-generated security assessments
    • Seamless integration with Outlook's existing interface
  6. Security Considerations:

    • Implemented input sanitization to prevent XSS attacks
    • Secured API communication with proper authentication and 100% local inference of Llama 3 70B
    • Ensured sensitive information (like API keys) is not exposed in the codebase
  7. Development Process:

    • Set up the initial project structure and development environment
    • Developed core functionality for email analysis
    • Created and styled the user interface
    • Integrated with the AI service for threat detection
    • Implemented theme switching and responsiveness
    • Conducted testing and debugging
    • Prepared the project for version control with a comprehensive .gitignore file
  8. Potential Impact:

    • Enhances email security for Outlook users and large enterpise users like banks or hospitals
    • Provides an accessible tool for identifying potential phishing or malicious emails
    • Educates users about email security through detailed explanations
  9. Future Possibilities:

    • Further refinement of the AI model for more accurate threat detection as there was limited time and not enough syntehtic data generated
    • Additional features like bulk email analysis or integration with other security tools
    • Massive potential for deployment in enterprise environments for enhanced organizational security

This project demonstrates a practical application of AI in everyday productivity tools, addressing the critical need for improved email security. By creating a user-friendly interface and integrating advanced AI capabilities, Vanguard stands as an innovative solution developed within the constraints of a hackathon, showcasing both technical skill and a keen understanding of user needs in the realm of digital security.

About

Vanguard is an Office 365 Add-in for advanced email scanning to prevent malicous attacks on an enterprise scale

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •