Threat Modeling Dataset for Software Systems

Threat Modeling Dataset for Software Systems Author: Leonid Zelenskiy Supervisor: Andrey Sadovykh Repository: github.com/RamPrin/LLM_TM Overview This repository contains a curated dataset for evaluating threat modeling approaches, specifically focused on cybersecurity testing and large language models (LLMs). It includes structured software system specifications and corresponding cybersecurity threats to support experiments in automated threat detection, analysis, and model evaluation. Dataset Structure test_dataset A CSV file containing cleaned system specifications suitable for natural language processing (NLP) tools. Format: <system, text> validation_dataset A CSV file listing manually identified cybersecurity threats used for validation. Format: <system, threat_title> Purpose This dataset supports research in automating the threat modeling process, with particular relevance to cybersecurity testing scenarios. It enables benchmarking of model performance in recognizing and generating relevant security threats based on structured system descriptions. Sources System information was extracted from a variety of reliable and public documentation, including RFCs, whitepapers, technical blogs, and academic research. Below is a mapping of each system to its original source. System Description Sources System Source OAuth 2.0 RFC 6819 SSL SSL Labs DNS Netmeister S3 TrustOnCloud Google Cloud Service NCC Group IoT Authentication SAFECode Whitepaper PCI DSS Shostack Paper Certificate Transparency IETF Draft Kubernetes (K8S) Link 1, Link 2, Link 3 CI/CD GitHub Threat Matrix AWS ECS Fargate Sysdig Password Store Manager Stanford Paper IoT Supply Chain ENISA Trinity Link Connected Cars Trend Micro Email Encryption Gateway NCC Group Bitcoin GitHub Containers Checklist, CloudSecDocs Medical Devices MITRE Contact Tracing Applications LinkedIn Article Vehicle Charging PNNL Agentic AI OWASP GenAI Trusted Firmware-M Trusted Firmware Docs ROS 2 Robotic System ROS2 Design Docs