Skip to content

coder7100/LLM_risk_analysis

 
 

Repository files navigation

Large Language Model (LLM) Risk Analysis Project

Overview

This repository is dedicated to the "LLM Risk Navigator" project by Team Patriots. Our goal is to develop an industry-specific risk matrix that helps organizations to prioritize mitigation efforts for LLM integration. This matrix is designed to empower businesses to navigate the complexities of LLM integration with confidence and resilience.

Project Goals

  • Security Assessments: To accompany the rising integration of LLMs into business systems with robust security assessments.
  • Mitigation Strategies: Development of tailored risk assessment methodologies to proactively address security concerns.
  • Industry-Specific Solutions: Innovating solutions that are tailored to the specific needs and vulnerabilities of various industries. Structure

The repository is structured as follows:

  • Markov Chain Analysis: This directory contains Markov Chain Analysis for each sector. It also has number of LLM attacks for each sector for each month.
  • News Scrapping: This directory includes the code as well as the data that we acquired from web scrapping of the news articles which are related to LLM attacks.
  • Statewise Threat Assesment: This directory includes the statewise analysis of LLM attacks. This data is used for tableau visualization which visualizes which state is more vulnerable to each sector.
  • Veris Dataset: This is the Veris Community Database (VCDB) which is a repository of cyber attacks happening in the whole world. We are using this dataset for our LLM attacks mapping. Here is the link of actual dataset: VERIS

Key Components

  • Threat Identification: Analysis of known attack vectors against LLMs.
  • Vulnerability Assessment: Using tools like Markov Chain analysis to predict potential attack paths and vulnerabilities.
  • Consequence Analysis: Evaluating the potential impacts of successful attacks on different sectors.

Technologies Used

  • Python: For data manipulation and analysis.
  • R: For statistical analysis and visualization.
  • Tableau: For creating interactive dashboards to visualize risks.

Team

  • Baba Akhil Sarvasiddhi - Product Owner
  • Davon Carvalho - Developer
  • Vishnu Vasireddy - Developer
  • Charan Reddy Bobbala - Developer
  • Aakash Aundhkar - Developer & Quality Manager
  • Susmith Poreddy - Scrum Master

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 100.0%