Skip to content

Latest commit

 

History

History
104 lines (87 loc) · 4.49 KB

README.md

File metadata and controls

104 lines (87 loc) · 4.49 KB

Data-Modeling-Group-2-Project

This is the Term Project for the Course entitled Data Modeling for Intelligent Networks and Security (ESW7002-41), Fall 2024.

Overview

This repository contains the implementation of an AI-driven solution to automate security policy generation. Using OpenAI's GPT-4o-mini language model, the system translates natural language descriptions into XML policies compliant with the I2NSF (Interface to Network Security Functions) Consumer-Facing Interface. This project aims to simplify network security configuration for non-technical users by automating the creation of machine-readable, standards-compliant security policies.

Features

  • Natural Language to XML Translation: Converts user-friendly policy descriptions into structured XML policies.
  • I2NSF Compliance: Generates policies following the I2NSF Consumer-Facing Interface schema.
  • Ease of Use: Enables non-technical users to define high-level security policies.
  • Event-Condition-Action Model: Supports defining events (system-events, system-alarms), conditions (e.g., time, URL categories) and actions (e.g., blocking or allowing traffic).

System Architecture

The implementation leverages the following key components:

  1. Natural Language Understanding: Processes user input to identify key components like events, conditions, and actions.
  2. Prompt Engineering: Utilizes a tailored prompt to guide the GPT-4o-mini model in generating XML outputs compliant with the I2NSF schema.
  3. YANG Data Model Integration: Incorporates the standardized Consumer-Facing Interface schema to ensure output interoperability and consistency.

System Diagram

Example

Input: "Block SNS access during office hours with a weekly frequency (9am-5pm; Monday-Friday)."

Generated Output:

<?xml version="1.0" encoding="UTF-8" ?>
<i2nsf-cfi-policy xmlns="urn:ietf:params:xml:ns:yang:ietf-i2nsf-cfi-policy">
    <name>block_sns_access_during_office_hours</name>
    <rules>
        <name>block_sns_access</name>
        <condition>
            <time>
                <frequency>weekly</frequency>
                <period>
                    <start-time>09:00</start-time>
                    <end-time>17:00</end-time>
                    <day>Monday</day>
                    <day>Tuesday</day>
                    <day>Wednesday</day>
                    <day>Thursday</day>
                    <day>Friday</day>
                </period>
            </time>
            <url-category>
                <url-name>SNS</url-name>
            </url-category>
        </condition>
        <actions>
            <primary-action>
                <action>drop</action>
            </primary-action>
        </actions>
    </rules>
</i2nsf-cfi-policy>

Getting Started

Prerequisites

  • Python 3.8+
  • OpenAI API Access
  • Jupyter Notebook (for running the provided .ipynb file)

Installation

  1. Clone the repository: 
    git clone https://github.com/jaehoonpauljeong/Data-Modeling-Group-2-Project
  2. Install dependencies: 
    pip install -r requirements.txt

Usage

IMPORTANT: This notebook has been implemented in Kaggle, modifications for different environments may be required.

  1. Open the Jupyter Notebook natural-language-to-security-policy.ipynb.
  2. Replace the OpenAI API key placeholder with your key.
  3. Input your desired policy descriptions and run the cells to generate XML policies.

Project Structure

  • natural-language-to-security-policy.ipynb: Main implementation notebook.
  • README.md: Documentation of the project.
  • examples/: Directory containing example input and output policies.
  • playground/fewshot_experiments.ipynb: Additional implementation using prompt selector.
  • requirements.txt: Python dependencies.

Contributing

Contributions are welcome! Please submit a pull request or open an issue for bug fixes, feature requests, or improvements.

Future Work

  • Expand datasets for quantitative evaluation of generated policies.
  • Enhance model scalability and context awareness for diverse cybersecurity scenarios.
  • Integrate the implementation with I2NSF systems for end-to-end functionality.

Authors

  • Mikel Larrarte Rodriguez, Jorge Alcorta Berasategui, Jaehoon (Paul) Jeong

License

This project is licensed under the MIT License.

References