awesome-MLSecOps
A curated list of MLSecOps tools, articles and other resources on security applied to Machine Learning and MLOps systems.
Stars: 204
Awesome MLSecOps is a curated list of open-source tools, resources, and tutorials for MLSecOps (Machine Learning Security Operations). It includes a wide range of security tools and libraries for protecting machine learning models against adversarial attacks, as well as resources for AI security, data anonymization, model security, and more. The repository aims to provide a comprehensive collection of tools and information to help users secure their machine learning systems and infrastructure.
README:
A curated list of awesome open-source tools, resources, and tutorials for MLSecOps (Machine Learning Security Operations).
- Open Source Security Tools
- Commercial Tools
- DATA
- ML Code Security
- 101 Resources
- Attack Vectors
- Blogs and Publications
- MLOps Infrastructure Vulnerabilities
- Community Resources
- Infographics
- Contributions
- Contributors
In this section, you and I can take a look at what opensource solutions and PoCs, exist to accomplish the task of ML protection. Of course, some of them are unsupported or will have difficulties to run. However, not mentioning them is a big crime.
| Tool | Description |
|---|---|
| ModelScan | Protection Against ML Model Serialization Attacks |
| NB Defense | Secure Jupyter Notebooks |
| Garak | LLM vulnerability scanner |
| Adversarial Robustness Toolbox | Library of defense methods for ML models against adversarial attacks |
| MLSploit | Cloud framework for interactive experimentation with adversarial machine learning research |
| TensorFlow Privacy | Library of privacy-preserving machine learning algorithms and tools |
| Foolbox | Python toolbox for creating and evaluating adversarial attacks and defenses |
| Advertorch | Python toolbox for adversarial robustness research |
| Artificial Intelligence Threat Matrix | Framework for identifying and mitigating threats to machine learning systems |
| Adversarial ML Threat Matrix | Adversarial Threat Landscape for AI Systems |
| CleverHans | A library of adversarial examples and defenses for machine learning models |
| AdvBox | Advbox is a toolbox to generate adversarial examples that fool neural networks in PaddlePaddle、PyTorch、Caffe2、MxNet、Keras、TensorFlow |
| Audit AI | Bias Testing for Generalized Machine Learning Applications |
| Deep Pwning | Deep-pwning is a lightweight framework for experimenting with machine learning models with the goal of evaluating their robustness against a motivated adversary |
| Privacy Meter | An open-source library to audit data privacy in statistical and machine learning algorithms |
| TensorFlow Model Analysis | A library for analyzing, validating, and monitoring machine learning models in production |
| PromptInject | A framework that assembles adversarial prompts |
| TextAttack | TextAttack is a Python framework for adversarial attacks, data augmentation, and model training in NLP |
| OpenAttack | An Open-Source Package for Textual Adversarial Attack |
| TextFooler | A Model for Natural Language Attack on Text Classification and Inference |
| Flawed Machine Learning Security | Practical examples of "Flawed Machine Learning Security" together with ML Security best practice across the end to end stages of the machine learning model lifecycle from training, to packaging, to deployment |
| Adversarial Machine Learning CTF | This repository is a CTF challenge, showing a security flaw in most (all?) common artificial neural networks. They are vulnerable for adversarial images |
| Damn Vulnerable LLM Project | A Large Language Model designed for getting hacked |
| Gandalf Lakera | Prompt Injection CTF playground |
| Vigil | LLM prompt injection and security scanner |
| PALLMs (Payloads for Attacking Large Language Models) | list of various payloads for attacking LLMs collected in one place |
| AI-exploits | exploits for MlOps systems. It's not just in the inputs given to LLMs such as ChatGPT |
| Offensive ML Playbook | Offensive ML Playbook. Notes on machine learning attacks and pentesting |
| AnonLLM | Anonymize Personally Identifiable Information (PII) for Large Language Model APIs |
| AI Goat | vulnerable LLM CTF challenges |
| Pyrit | The Python Risk Identification Tool for generative AI |
| Raze to the Ground: Query-Efficient Adversarial HTML Attacks on Machine-Learning Phishing Webpage Detectors | Source code of the paper "Raze to the Ground: Query-Efficient Adversarial HTML Attacks on Machine-Learning Phishing Webpage Detectors" accepted at AISec '23 |
| Giskard | Open-source testing tool for LLM applications |
| Safetensors | Convert pickle to a safe serialization option |
| Citadel Lens | Quality testing of models according to industry standards |
| Model-Inversion-Attack-ToolBox | A framework for implementing Model Inversion attacks |
| NeMo-Guardials | NeMo Guardrails allow developers building LLM-based applications to easily add programmable guardrails between the application code and the LLM |
| AugLy | A tool for generating adversarial attacks |
| Knockoffnets | PoC to implement BlackBox attacks to steal model data |
| Robust Intelligence Continous Validation | Tool for continuous model validation for compliance with standards |
| VGER | Jupyter Attack framework |
| AIShield Watchtower | An open source tool from AIShield for studying AI models and scanning for vulnerabilities |
| PS-fuzz | tool for scanning LLM vulnerabilities |
| Mindgard-cli | Check security of you AI via CLI |
| PurpleLLama3 | Check LLM security with Meta LLM Benchmark |
| Model transparency | generate model signing |
| ARTkit | Automated prompt-based testing and evaluation of Gen AI applications |
| LangBiTe | A Bias Tester framework for LLMs |
| OpenDP | The core library of differential privacy algorithms powering the OpenDP Project |
| TF-encrypted | Encryption for tensorflow |
| Tool | Description |
|---|---|
| Databricks Platform, Azure Databricks | Datalake data management and implementation tool |
| Hidden Layer AI Detection Response | Tool for detecting and responding to incidents |
| Guardian | Model protection in CI/CD |
| Tool | Description |
|---|---|
| ARX - Data Anonymization Tool | Tool for anonymizing datasets |
| Data-Veil | Data masking and anonymization tool |
| Tool for IMG anonymization | Image anonymization |
| Tool for DATA anonymization | Data anonymization |
| BMW-Anonymization-Api | This repository allows you to anonymize sensitive information in images/videos. The solution is fully compatible with the DL-based training/inference solutions that we already published/will publish for Object Detection and Semantic Segmentation |
| DeepPrivacy2 | A Toolbox for Realistic Image Anonymization |
| PPAP | Latent-space-level Image Anonymization with Adversarial Protector Networks |
- lintML - Security linter for ML, by Nvidia
- HiddenLayer: Model as Code - Research about some vectors in ML libraries
- Copycat CNN - Proof-of-concept on how to generate a copy of a Convolutional Neural Network
- differential-privacy-library - Library designed for differential privacy and machine learning
You can find here a list of resources to help you get into the topic of AI security. Understand what attacks exist and how they can be used by an attacker.
- AI Security 101
- Web LLM attacks
- Microsoft AI Red Team
- AI Risk Assessment for ML Engineers
- Microsoft - Generative AI Security for beginners
Full size map in this repository
more in Adversarial AI Attacks, Mitigations, and Defense Strategies: A cybersecurity professional's guide to AI attacks, threat modeling, and securing AI with MLSecOps
🌱 The AI security community is growing. New blogs and many researchers are emerging. In this paragraph you can see examples of some blogs.
- 🛡️ Red-Teaming Large Language Models
- 🔍 Google's AI red-team
- 🔒 The MLSecOps Top 10 vulnerabilities
- 🏴☠️ Token Smuggling Jailbreak via Adversarial Prompt
- ☣️ Just How Toxic is Data Poisoning? A Unified Benchmark for Backdoor and Data Poisoning Attacks
- 📊 We need a new way to measure AI security
- 🕵️ PrivacyRaven: Implementing a proof of concept for model inversion
- 🧠 Adversarial Prompts Engineering
- 🔫 TextAttack: A Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLP
- 📋 Trail Of Bits' audit of Hugging Face's safetensors library
- 🔝 OWASP Top 10 for Large Language Model Applications
- 🔐 LLM Security
- 🔑 Is you MLOps infrastructure leaking secrets?
- 🚩 Embrace The Red, blog where show how u can hack LLM's.
- 🎙️ Audio-jacking: Using generative AI to distort live audio transactions
- 🌐 HADESS - Web LLM Attacks
- 🧰 WTF-blog - MlSecOps frameworks ... Which ones are available and what is the difference?
- 📚 DreadNode Paper Stack
Very interesting articles on MlOps infrastructure vulnerabilities. In some of them you can even find ready-made exploits.
- SILENT SABOTAGE - Study on bot compromise for converting Pickle to SafeTensors
- NOT SO CLEAR: HOW MLOPS SOLUTIONS CAN MUDDY THE WATERS OF YOUR SUPPLY CHAIN - Study on vulnerabilities for the ClearML platform
- Uncovering Azure's Silent Threats: A Journey into Cloud Vulnerabilities - Study on security issues of Azure MLAAS
- The MLOps Security Landscape
- Confused Learning: Supply Chain Attacks through Machine Learning Models
Official implementation of "AgentPoison: Red-teaming LLM Agents via Memory or Knowledge Base Backdoor Poisoning". This project explores methods of data poisoning and backdoor insertion in LLM agents to assess their resilience against such attacks.
Research on methods of embedding malicious payloads into deep neural networks.
Investigation of backdoor attacks on deep learning models, focusing on creating undetectable vulnerabilities within models.
Techniques for stealing deep learning models through various attack vectors, enabling adversaries to replicate or access models.
Model extraction without using data, allowing for the recovery of models without access to the original data.
Tool for mapping and analyzing large language models (LLMs), exploring the structure and behavior of various LLMs.
Federated learning pipeline using Google Cloud infrastructure, enabling model training on distributed data.
Attack using ensemble class activation maps to introduce errors in models by manipulating activation maps.
Methods for attacking deep models under various conditions and constraints, focusing on creating more resilient attacks.
Research on adaptive attacks on machine learning models, enabling the creation of attacks that can adapt to model defenses.
Knowledge transfer in zero-shot scenarios, exploring methods to transfer knowledge between models without prior training on target data.
Attack for generating informative labels, aimed at covertly extracting data from trained models.
Enhancing DMI (Data Mining and Integration) methods using additional knowledge to improve accuracy and efficiency.
Research on methods for visualizing and interpreting machine learning models, providing insights into model workings.
Attacks that can be "plugged and played" without needing model modifications, offering flexible and universal attack methods.
Tool for analyzing and processing snapshot data, enabling efficient handling of data snapshots.
Research on the trade-offs between privacy and robustness in models, aiming to balance these two aspects in machine learning.
Methods for data leakage from trained models, exploring ways to extract private information from machine learning models.
Research on blind information extraction attacks, enabling data retrieval without access to the model's internal structure.
Differential privacy methods for deep learning, ensuring data privacy during model training.
Defense methods using MMD-mixup, aimed at improving model robustness against attacks.
Tools for protecting memory from attacks, exploring ways to prevent data leaks from model memory.
Methods for merging and splitting data to improve training, optimizing the use of heterogeneous data in models.
Attacks on face recognition models using attributes, exploring ways to manipulate facial attributes to induce errors.
Attacks on face verification models, aimed at disrupting authentication systems based on face recognition.
Using GANs to create malware, exploring methods for generating malicious code with generative models.
Methods for generating adversarial perturbations using generative models, aimed at introducing errors in deep models.
Adversarial attacks using Relativistic AdvGAN, exploring methods for creating more realistic and effective attacks.
Attacks on large language models, exploring vulnerabilities and protection methods for LLMs.
Safe fine-tuning of large language models, aiming to prevent data leaks and ensure security during LLM tuning.
Methods for evaluating trust in models, exploring ways to determine the reliability and safety of machine learning models.
Benchmark for evaluating prompts, providing tools for testing and optimizing queries to large language models.
Tool for analyzing and evaluating models based on ROM codes, exploring various aspects of model performance and resilience.
Research on privacy in large language models, aiming to protect data and prevent leaks from LLMs.
- MLSecOps
- MLSecOps Podcast
- MITRE ATLAS™ and SLACK COMMUNITY
- MlSecOps communtiy and SLACK COMMUNITY
- MITRE ATLAS™ (Adversarial Threat Landscape for Artificial-Intelligence Systems)
- OWASP AI Exchange
- OWASP Machine Learning Security Top Ten
- OWASP Top 10 for Large Language Model Applications
- OWASP LLMSVS
- OWASP Periodic Table of AI Security
- OWASP SLACK
- Awesome LLM Security
- Hackstery
- PWNAI
- AiSec_X_Feed
- HUNTR Discord community
- AIRSK
- AI Vulnerability Database
- Incident AI Database
- Defcon AI Village CTF
- Awesome AI Security
- MLSecOps Reference Repository
- Awesome LVLM Attack
- Awesome MLLM Safety
- Adversarial AI Attacks, Mitigations, and Defense Strategies: A cybersecurity professional's guide to AI attacks, threat modeling, and securing AI with MLSecOps
- Privacy-Preserving Machine Learning
- Generative AI Security: Theories and Practices (Future of Business and Finance)
All contributions to this list are welcome! Please feel free to submit a pull request with any additions or improvements.
 @riccardobiosas |
 @badarahmed |
 @deadbits |
 @wearetyomsmnv |
 @anmorgan24 |
 @mik0w |
 @alexcombessie |
If you find this project useful, please consider giving it a star ⭐️
This project is licensed under the MIT License - see the LICENSE file for details.
Made with ❤️
For Tasks:
Click tags to check more tools for each tasksFor Jobs:
Alternative AI tools for awesome-MLSecOps
Similar Open Source Tools
awesome-MLSecOps
Awesome MLSecOps is a curated list of open-source tools, resources, and tutorials for MLSecOps (Machine Learning Security Operations). It includes a wide range of security tools and libraries for protecting machine learning models against adversarial attacks, as well as resources for AI security, data anonymization, model security, and more. The repository aims to provide a comprehensive collection of tools and information to help users secure their machine learning systems and infrastructure.
inference
Xorbits Inference (Xinference) is a powerful and versatile library designed to serve language, speech recognition, and multimodal models. With Xorbits Inference, you can effortlessly deploy and serve your or state-of-the-art built-in models using just a single command. Whether you are a researcher, developer, or data scientist, Xorbits Inference empowers you to unleash the full potential of cutting-edge AI models.
SimAI
SimAI is the industry's first full-stack, high-precision simulator for AI large-scale training. It provides detailed modeling and simulation of the entire LLM training process, encompassing framework, collective communication, network layers, and more. This comprehensive approach offers end-to-end performance data, enabling researchers to analyze training process details, evaluate time consumption of AI tasks under specific conditions, and assess performance gains from various algorithmic optimizations.
camel
CAMEL is an open-source library designed for the study of autonomous and communicative agents. We believe that studying these agents on a large scale offers valuable insights into their behaviors, capabilities, and potential risks. To facilitate research in this field, we implement and support various types of agents, tasks, prompts, models, and simulated environments.
AIOS
AIOS, a Large Language Model (LLM) Agent operating system, embeds large language model into Operating Systems (OS) as the brain of the OS, enabling an operating system "with soul" -- an important step towards AGI. AIOS is designed to optimize resource allocation, facilitate context switch across agents, enable concurrent execution of agents, provide tool service for agents, maintain access control for agents, and provide a rich set of toolkits for LLM Agent developers.
Academic_LLM_Sec_Papers
Academic_LLM_Sec_Papers is a curated collection of academic papers related to LLM Security Application. The repository includes papers sorted by conference name and published year, covering topics such as large language models for blockchain security, software engineering, machine learning, and more. Developers and researchers are welcome to contribute additional published papers to the list. The repository also provides information on listed conferences and journals related to security, networking, software engineering, and cryptography. The papers cover a wide range of topics including privacy risks, ethical concerns, vulnerabilities, threat modeling, code analysis, fuzzing, and more.
ai-agents-for-beginners
AI Agents for Beginners is a course that covers the fundamentals of building AI Agents. It consists of 10 lessons with code examples using Azure AI Foundry and GitHub Model Catalogs. The course utilizes AI Agent frameworks and services from Microsoft, such as Azure AI Agent Service, Semantic Kernel, and AutoGen. Learners can access written lessons, Python code samples, and additional learning resources for each lesson. The course encourages contributions and suggestions from the community and provides multi-language support for learners worldwide.
cambrian
Cambrian-1 is a fully open project focused on exploring multimodal Large Language Models (LLMs) with a vision-centric approach. It offers competitive performance across various benchmarks with models at different parameter levels. The project includes training configurations, model weights, instruction tuning data, and evaluation details. Users can interact with Cambrian-1 through a Gradio web interface for inference. The project is inspired by LLaVA and incorporates contributions from Vicuna, LLaMA, and Yi. Cambrian-1 is licensed under Apache 2.0 and utilizes datasets and checkpoints subject to their respective original licenses.
LLaVA-MORE
LLaVA-MORE is a new family of Multimodal Language Models (MLLMs) that integrates recent language models with diverse visual backbones. The repository provides a unified training protocol for fair comparisons across all architectures and releases training code and scripts for distributed training. It aims to enhance Multimodal LLM performance and offers various models for different tasks. Users can explore different visual backbones like SigLIP and methods for managing image resolutions (S2) to improve the connection between images and language. The repository is a starting point for expanding the study of Multimodal LLMs and enhancing new features in the field.
openrl
OpenRL is an open-source general reinforcement learning research framework that supports training for various tasks such as single-agent, multi-agent, offline RL, self-play, and natural language. Developed based on PyTorch, the goal of OpenRL is to provide a simple-to-use, flexible, efficient and sustainable platform for the reinforcement learning research community. It supports a universal interface for all tasks/environments, single-agent and multi-agent tasks, offline RL training with expert dataset, self-play training, reinforcement learning training for natural language tasks, DeepSpeed, Arena for evaluation, importing models and datasets from Hugging Face, user-defined environments, models, and datasets, gymnasium environments, callbacks, visualization tools, unit testing, and code coverage testing. It also supports various algorithms like PPO, DQN, SAC, and environments like Gymnasium, MuJoCo, Atari, and more.
buffer-of-thought-llm
Buffer of Thoughts (BoT) is a thought-augmented reasoning framework designed to enhance the accuracy, efficiency, and robustness of large language models (LLMs). It introduces a meta-buffer to store high-level thought-templates distilled from problem-solving processes, enabling adaptive reasoning for efficient problem-solving. The framework includes a buffer-manager to dynamically update the meta-buffer, ensuring scalability and stability. BoT achieves significant performance improvements on reasoning-intensive tasks and demonstrates superior generalization ability and robustness while being cost-effective compared to other methods.
airunner
AI Runner is a multi-modal AI interface that allows users to run open-source large language models and AI image generators on their own hardware. The tool provides features such as voice-based chatbot conversations, text-to-speech, speech-to-text, vision-to-text, text generation with large language models, image generation capabilities, image manipulation tools, utility functions, and more. It aims to provide a stable and user-friendly experience with security updates, a new UI, and a streamlined installation process. The application is designed to run offline on users' hardware without relying on a web server, offering a smooth and responsive user experience.
AReaL
AReaL (Ant Reasoning RL) is an open-source reinforcement learning system developed at the RL Lab, Ant Research. It is designed for training Large Reasoning Models (LRMs) in a fully open and inclusive manner. AReaL provides reproducible experiments for 1.5B and 7B LRMs, showcasing its scalability and performance across diverse computational budgets. The system follows an iterative training process to enhance model performance, with a focus on mathematical reasoning tasks. AReaL is equipped to adapt to different computational resource settings, enabling users to easily configure and launch training trials. Future plans include support for advanced models, optimizations for distributed training, and exploring research topics to enhance LRMs' reasoning capabilities.
awesome-flux-ai
Awesome Flux AI is a curated list of resources, tools, libraries, and applications related to Flux AI technology. It serves as a comprehensive collection for developers, researchers, and enthusiasts interested in Flux AI. The platform offers open-source text-to-image AI models developed by Black Forest Labs, aiming to advance generative deep learning models for media, creativity, efficiency, and diversity.
mage-ai
Mage is an open-source data pipeline tool for transforming and integrating data. It offers an easy developer experience, engineering best practices built-in, and data as a first-class citizen. Mage makes it easy to build, preview, and launch data pipelines, and provides observability and scaling capabilities. It supports data integrations, streaming pipelines, and dbt integration.
superduperdb
SuperDuperDB is a Python framework for integrating AI models, APIs, and vector search engines directly with your existing databases, including hosting of your own models, streaming inference and scalable model training/fine-tuning. Build, deploy and manage any AI application without the need for complex pipelines, infrastructure as well as specialized vector databases, and moving our data there, by integrating AI at your data's source: - Generative AI, LLMs, RAG, vector search - Standard machine learning use-cases (classification, segmentation, regression, forecasting recommendation etc.) - Custom AI use-cases involving specialized models - Even the most complex applications/workflows in which different models work together SuperDuperDB is **not** a database. Think `db = superduper(db)`: SuperDuperDB transforms your databases into an intelligent platform that allows you to leverage the full AI and Python ecosystem. A single development and deployment environment for all your AI applications in one place, fully scalable and easy to manage.
For similar tasks
awesome-MLSecOps
Awesome MLSecOps is a curated list of open-source tools, resources, and tutorials for MLSecOps (Machine Learning Security Operations). It includes a wide range of security tools and libraries for protecting machine learning models against adversarial attacks, as well as resources for AI security, data anonymization, model security, and more. The repository aims to provide a comprehensive collection of tools and information to help users secure their machine learning systems and infrastructure.
For similar jobs
awesome-MLSecOps
Awesome MLSecOps is a curated list of open-source tools, resources, and tutorials for MLSecOps (Machine Learning Security Operations). It includes a wide range of security tools and libraries for protecting machine learning models against adversarial attacks, as well as resources for AI security, data anonymization, model security, and more. The repository aims to provide a comprehensive collection of tools and information to help users secure their machine learning systems and infrastructure.
mimir
MIMIR is a Python package designed for measuring memorization in Large Language Models (LLMs). It provides functionalities for conducting experiments related to membership inference attacks on LLMs. The package includes implementations of various attacks such as Likelihood, Reference-based, Zlib Entropy, Neighborhood, Min-K% Prob, Min-K%++, Gradient Norm, and allows users to extend it by adding their own datasets and attacks.
openshield
OpenShield is a firewall designed for AI models to protect against various attacks such as prompt injection, insecure output handling, training data poisoning, model denial of service, supply chain vulnerabilities, sensitive information disclosure, insecure plugin design, excessive agency granting, overreliance, and model theft. It provides rate limiting, content filtering, and keyword filtering for AI models. The tool acts as a transparent proxy between AI models and clients, allowing users to set custom rate limits for OpenAI endpoints and perform tokenizer calculations for OpenAI models. OpenShield also supports Python and LLM based rules, with upcoming features including rate limiting per user and model, prompts manager, content filtering, keyword filtering based on LLM/Vector models, OpenMeter integration, and VectorDB integration. The tool requires an OpenAI API key, Postgres, and Redis for operation.
paig
PAIG is an open-source project focused on protecting Generative AI applications by ensuring security, safety, and observability. It offers a versatile framework to address the latest security challenges and integrate point security solutions without rewriting applications. The project aims to provide a secure environment for developing and deploying GenAI applications.