Threat Model-driven Test Framework for Security and Privacy of Agentic LLM Applications: supporting materials

This repository provides the supporting materials for the submission entitled "A Security and Privacy Threat Model for Agentic LLM Applications".

It consists of 

Threat model

ThreatModel.pdf: a self-contained and complete threat model report, which includes:

DFD and details about the DFD construction

STRIDE and LINDDUN analysis and results

In-depth mapping to other catalogs and frameworks, i.e. OWASP, MCP, and Unit41.

Threat model source files (SPARTA)

Raw CSV of the generated threats

Completed threat model Excel file which contains the threats, their severity assessment, the mapping onto frameworks and onto the test cases

Test framework

README.md / README.pdf providing detailed explanations of the test framework design, test case definition, and instructions to execute and reproduce

Source code of the test framework, 

Source code of the implemented test cases,

Logs of the reported test runs and their detailed outcomes