Skip to main content
Back to top
Ctrl
+
K
PyRIT
Cookbooks
1. Sending a Million Prompts
2. Precomputing Turns for Attacks
Probing for copyright violations with FirstLetterConverter and PlagiarismScorer
4. Benchmarking models + Testing For Fairness/Bias
Install PyRIT Library
Install PyRIT with Docker
What can I do if Jupyter cannot find PyRIT?
Populating Secrets - Quick Start Guide
Azure SQL Database Setup, Authentication and User Permissions
Contributing
1. Local Installation with Conda/Python
DevContainers Setup in Visual Studio Code
2. Contribute with Git
3. Incorporating Research
4. Style Guide
5. Running Tests
6. Unit Tests
7. Integration Tests
8. Notebooks
9. Pre-Commit Hooks
10. Exception Handling Guidelines
11. Releasing PyRIT
Architecture
User guide
Datasets
1. Seed Prompts
2. Fetching Datasets
Executor
Attack
1. Prompt Sending Attack (Single-Turn)
2. Red Teaming Attack (Multi-Turn)
3. Crescendo Attack (Multi-Turn)
Skeleton Key Attack (Single-Turn) - optional
Violent Durian Attack (Multi-Turn) - optional
Flip Attack (Single-Turn) - optional
Context Compliance Attack (Single-Turn) - optional
Role Play Attack (Single-Turn) - optional
Many-Shot Jailbreak Attack (Single-Turn) - optional
Tree of Attacks with Pruning (Multi-Turn) - optional
Multi-Prompt Sending Attack - optional
Workflow
1. Cross-domain Prompt Injection Attack (XPIA) via a website
2. Cross-domain Prompt Injection (XPIA) - AI Recruiter RAG Vulnerability
Benchmark
1. Q&A Benchmark
Prompt Generator
1. Anecdoctor Prompt Generator
GPTFuzzer Prompt Generator - optional
Prompt Targets
1. OpenAI Chat Target
2. Creating Custom Targets
3. AML Chat Targets
4. Azure Blob Storage Targets
5. Multi-Modal Targets
6. Rate Limit (RPM) Threshold
7. HTTP Target
8. OpenAI Responses Target
OpenAI Completions - optional
Playwright Target - optional
Prompt Shield Target - optional
HuggingFace Chat Target - optional
Realtime Target - optional
Converters
1. Converters with LLMs
2. Using Prompt Converters
3. Audio Converters
4. Image Converters
5. Selectively Converting
6. Human in the Loop Converter
7. Video Converters
AnsiAttackConverter - optional
Generating Perturbed Prompts Using the CharSwapConverter - optional
PDFConverter - optional
MathPromptConverter - optional
Transparency Attack Converter: Crafting Images with Imperceptible Layers
Scoring
1. Float Scale Scoring using Azure Content Safety API
2. True False Scoring
3. Classification Scoring
4. Float Scale Scoring using Likert Scale
5. Human in the Loop Scoring with Gradio
6. Refusal Scorer
7. Batch Scoring
Insecure Code Scorer - optional
LookBack Scorer - optional
Prompt Shield Scorer - optional
Generic Self-Ask Scorer - optional
Scorer Evaluations - optional
Memory
1. SQLite Memory
2. Basic Memory Programming Usage
3. Memory Types
4. Working with Memory Manually
5. Resending Prompts Using Memory Labels Example
6. Azure SQL Memory
7. PromptSendingAttack with Azure SQL Memory
8. Seed Prompt Database
9. Exporting Data Example
10. Memory Schema Diagram
Querying by Harm Categories
Azure OpenAI Embeddings - optional
Chat messages - optional
Configuring PyRIT
Auxiliary Attacks
1. Generating GCG Suffixes Using Azure Machine Learning
Scenarios
PyRIT Scanner
Scanner Configuration Reference
Deployments
Deploying Hugging Face Models into Azure ML Managed Online Endpoint
Importing and Registering Hugging Face Models into Azure ML
Hugging Face LLMs on Azure ML: Endpoint Interaction Guide
Score Azure ML Managed Online Endpoint
Troubleshooting Guide for HF Azure ML Models
API Reference
Blog
When External Data Becomes a Trojan Horse
A More Generalized OpenAIChatTarget
Datasets and Seed Prompts
Proxying PyRIT
Using PyRIT as a Bug Hunter
Multi-Turn orchestrators
Repository
Suggest edit
Open issue
.md
.pdf
Deployments
Deployments
#
