AI-Powered Threat Modeling for Secure System Design¶

Overview

This section covers AI-powered threat modeling for secure system design, focusing on open-source tools and prompt-based solutions. Below is a comparison of features across notable open-source tools that assist in early design-time threat modeling.

Threat Modeling¶

Quote

The best use of threat modeling is to improve the security and privacy of a system through early and frequent analysis.
Threat modeling must align with an organization’s development practices and follow design changes in iterations that are each scoped to manageable portions of the system.
The outcomes of threat modeling are meaningful when they are of value to stakeholders.
Dialog is key to establishing the common understandings that lead to value, while documents record those understandings, and enable measurement.

Threat Modeling Manifesto

Quote

Threat modeling is an investment—in my view, it’s a good one, because finding and mitigating threats in the design phase of your workload feature can reduce the relative cost of mitigation, compared to finding the threats later. Consistently implementing threat modeling will likely also improve your security posture over time.

https://aws.amazon.com/blogs/security/threat-modeling-your-generative-ai-workload-to-evaluate-security-risk/

Open Source Tools Comparison¶

Comparison of features and capabilities of open-source AI-driven threat modeling tools (focused on early design-phase usage, except AI Security Analyzer which also supports code-level analysis).

Tool	Focus & Approach	Key Features	AI / Model Integration
AWS Threat Designer	Design-phase automated threat modeling; Web UI (AWS Cloud stack). Users upload system architecture info (including diagrams) to generate threat models.	– Architecture diagram analysis: Multi-modal LLM analyzes system diagrams to identify components and relationships. – Threat catalog: Generates a comprehensive list of potential threats with interactive filtering and refinement. – Iterative replay: Allows rerunning the model after design changes to see updated risk postures. – Exports & history: Supports exporting results to PDF/DOCX and browsing past models in a catalog.	Uses AWS Bedrock with large models (Anthropic Claude 4 Sonnet by default) for NLP and vision; serverless backend. Multimodal LLM interprets text and diagrams to generate threats.
AWS Threat Composer	Design-phase guided threat modeling; Web app or VS Code extension. Emphasizes manual brainstorming with structured guidance (“threat model as code”).	– Data capture: Records system description, architecture and dataflow diagrams, assumptions, etc., as part of the model. – Threat grammar: Uses a prescriptive template to iteratively craft threat statements (with examples and suggestions) instead of starting from scratch. – Insights dashboard: Highlights gaps like threats without mitigations, unprioritized threats, category coverage, etc., to answer “did we do enough?”. – Packs & export: Supports reusable threat/mitigation packs for bulk addition (in self-hosted mode), and can generate a consolidated threat model document.	No built-in generative model – focuses on human-driven input with static guidance. (Uses local storage; no data leaves the browser. “Threat Grammar” framework provides structure rather than AI generation.)
StrideGPT	Design-phase automated STRIDE threat modeling; Streamlit web app or Docker CLI. Given an app description (and optionally diagrams or repo link), it auto-generates a full threat model.	– STRIDE threats generation: Uses LLM to output threats categorized by STRIDE (Spoofing, Tampering, etc.) for the provided system description. – Risk & mitigations: Automatically provides DREAD risk scores and suggests mitigation steps for each identified threat. – Attack trees & tests: Produces potential attack paths (attack trees) and even Gherkin-style security test cases based on threats. – Multi-modal input: Can accept architecture diagrams or flowcharts for analysis with vision-capable models; also can analyze a project’s repository (e.g. README) to enrich the threat model. – No data retention: Does not store inputs or results on the server, focusing on privacy.	Supports multiple LLMs via API or local runtime: OpenAI GPT-4 (and newer GPT-4.1), Anthropic Claude 3/4, Google Gemini (2.0/2.5) and others. Also compatible with self-hosted local models (Ollama, LM Studio). This flexible backend allows using the best available model for analysis.
AI Security Analyzer	Code-centered security analyzer with threat modeling output; CLI tool (Python) that scans an existing codebase to produce security documentation. Geared toward integrating threat modeling in later stages (after code exists) as well as design review.	– Multi-faceted analysis: Generates a Security Design Review document which includes threat modeling, attack surface analysis, attack trees, mitigation strategies, and identified vulnerabilities. – Code-aware threat identification: Parses project source code (multiple languages: Python, Java, JavaScript, Go, Android, etc.) to find assets, entry points, and potential threats specific to the implementation. – Automated documentation: Outputs comprehensive Markdown reports (e.g. `security_design.md`) that consolidate the security findings and model. – Flexible deployment: Can run via Python (Poetry) or Docker; cross-platform support (Windows, MacOS, Linux).	Leverages LLMs to analyze code and generate text. Supports multiple model providers: OpenAI API (GPT-4 family), Anthropic (Claude), Google PaLM/Gemini via API, and OpenRouter. The user supplies an API key for the chosen model, and the tool orchestrates prompts to produce the security docs.

Open Source Tools Details¶

AWS Threat Designer¶

AWS Threat Designer: AI-powered threat modeling for secure system design.

See blogpost: Accelerate threat modeling with generative AI for an in-depth overview of the solution.

Quote

Effective threat modeling examines data flows, trust boundaries, and potential attack vectors to create a comprehensive security strategy tailored to the specific system.

In a shift-left approach to security, threat modeling serves as a critical early intervention. By implementing threat modeling during the design phase—before a single line of code is written—organizations can identify and address potential vulnerabilities at their inception point.

AWS Accelerate threat modeling with generative AI, JUN 2025

Each function generates specialized prompts for different phases of the threat modeling process, including:

Asset identification
Data flow analysis
Gap analysis
Threat identification and improvement
Response structuring

https://github.com/awslabs/threat-designer/tree/main?tab=readme-ov-file#prerequisites

The backend is written in python: threat-designer/backend/threat_designer/

The associated prompts are https://github.com/awslabs/threat-designer/blob/0554b6a97c08e38bb92504ba13768780adb0301f/backend/threat_designer/prompts.py

AWS Threat Composer¶

A simple threat modeling tool to help humans to reduce time-to-value when threat modeling https://github.com/awslabs/threat-composer#readme

STRIDE GPT¶

Quote

Features:

Simple and user-friendly interface
Generates threat models based on the STRIDE methodology
Multi-modal: Use architecture diagrams, flowcharts, etc. as inputs for threat modelling across all supported vision-capable models
Generates attack trees to enumerate possible attack paths
Suggests possible mitigations for identified threats
Supports DREAD risk scoring for identified threats
Generates Gherkin test cases based on identified threats
GitHub repository analysis for comprehensive threat modelling
No data storage; application details are not saved
Supports models accessed via OpenAI API, Azure OpenAI Service, Google AI API, Mistral API, or locally hosted models via Ollama and 🆕 LM Studio Server
Available as a Docker container image for easy deployment
Environment variable support for secure configuration

https://github.com/mrwadams/stride-gpt

The app https://stridegpt.streamlit.app/ has these tabs

It supports a Bring Your Own LLM Key i.e. you chose the LLM and provide your API key.

AI Security Analyzer¶

Quote

AI Security Analyzer is a Python-based tool that analyzes your project's codebase and automatically generates detailed security documentation. It supports multiple analysis types:

🔒 Security Design Documentation
🎯 Threat Modeling
🔍 Attack Surface Analysis
🌳 Attack Tree Analysis
🛡️ Mitigation Strategies
🐛 Vulnerabilities

https://github.com/xvnpw/ai-security-analyzer

generate four different types of security documents:

🔒 Security Design Documentation: Generating detailed security design review.
🎯 Threat Modeling: Performing threat modeling analysis.
🔍 Attack Surface Analysis: Identifying potential entry points and vulnerabilities in the project’s attack surface.
🌳 Attack Tree Analysis: Visualizing potential attack vectors and their hierarchies through attack tree.

https://xvnpw.github.io/posts/scaling-threat-modeling-with-ai/

The associated prompts are

https://github.com/xvnpw/ai-security-analyzer/blob/dabfc57b6e5da9d99b3df5229fd496a224dac862/ai_security_analyzer/prompts.py

Additional Prompt Resources¶

Beyond full tools, there are also open-source prompt libraries focused on security threat modeling:

Fabric by Daniel Miessler: A crowdsourced collection of AI prompt “patterns.” It includes a create_stride_threat_model pattern to guide an LLM in producing a STRIDE-based threat model from a system description and a create_threat_scenarios pattern for generating detailed attack scenarios. These patterns can be used with various GPT-based systems to jump-start threat modeling exercises.
Kornelius Security Audit Prompt: An open prompt (from the Kornelius project) that provides a template for security auditing via LLMs. This prompt script can be adapted to evaluate a system’s security posture by enumerating threats and checks.

Each of the above resources provides ready-made prompt structures that practitioners can use with their AI of choice to conduct threat modeling, complementing the dedicated tools compared in the table.