Skip to content

Semih702/mcp-security-lab

Repository files navigation

MCP Security Lab

This repository contains the working materials for a final-year project focused on security weaknesses in MCP-based and LLM-connected systems.

The project direction is:

  • reproduce realistic security attack scenarios in a controlled lab
  • study why those attacks succeed
  • evaluate prevention and mitigation strategies
  • document layered threat models and concrete attack patterns

Main Contents

  • mcp_attack_lab/
    • Python-based lab scaffold for scenario definitions, MCP profiles, model configs, fixtures, and evaluation logic
  • mcp_security_threat_taxonomy_layered_restored.html
    • layered MCP threat taxonomy document
  • mcp_attack_catalog.html
    • attack-pattern catalog with concrete examples and relationships
  • mcp_runtime_defenses.md
    • notes on runtime defenses and control points
  • policy_gateway.md
    • notes on policy and enforcement ideas

Current State

The repository currently provides:

  • a reusable scenario schema
  • initial attack lab structure
  • first baseline scenario definition
  • research and documentation artifacts around MCP security

The next implementation phase is to connect the scenario runner to real MCP server execution, model adapters, trace capture, and pass/fail evaluation.

About

No description, website, or topics provided.

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors