The Bear Series is designed to introduce you to foundational AI security concepts. To begin, in the navigation bar, select Challenges > Series. Then, scroll down and select Bear. Through a series of engaging tasks and step-by-step walkthrough notebooks, you’ll dive into topics like data analysis, model extraction, and LLM manipulation. To complete these challenges, you’ll need:
  • The CHALLENGE_URL in the challenge instructions. For example: https://bear1.platform.dreadnode.io.
  • Your free DREADNODE_API_KEY. The API Key is located on each challenge page and in your Settings. If you need to regenerate your API Key, navigate to the /account page.

​
Get Started

Start by exploring each notebook for a detailed, step-by-step walkthrough. Whether you’re new to AI security or looking to expand your skills, the Bear Series will guide you through practical, real-world applications of cutting-edge AI security techniques. Let’s dive in and complete your first challenge!

​
Bear 1: Data Exploration

In this challenge, you’ll dive into API interaction and data analysis using pandas, a powerful Python library for handling data.

​
What You’ll Learn

  • Gain practical experience with pandas to manipulate and explore datasets.
  • Apply basic data analysis techniques to identify vulnerabilities and potential threats.
  • Learn to extract hidden information from data, a key skill in AI security and cybersecurity research.

​
Resources to Get Started

By the end of this challenge, you’ll be comfortable navigating datasets and using data analysis techniques to uncover critical insights, which are essential in cybersecurity.

​
Bear 2: Model Extraction

In this challenge, you’ll explore model extraction, a technique that helps you infer the internal workings of a machine learning model based on its outputs.

​
What You’ll Learn

  • How to replicate or analyze a model without having direct access to its internal structure.
  • Why model extraction is crucial for ensuring transparency and fairness in machine learning applications.
  • Key methods to analyze and understand how input data transforms into predictions.
This challenge will help you gain a deep understanding of how models behave and how to approach reverse-engineering to uncover vulnerabilities.

​
Bear 3: Model Fingerprinting

Here, you’ll apply your AI security skills to model fingerprinting.

​
What You’ll Learn

  • How to identify and characterize machine learning models based on their outputs.
  • Techniques for creating a “fingerprint” that distinguishes one model from another.
  • Why understanding a model’s architecture and preprocessing steps is essential for security professionals to anticipate potential vulnerabilities.
By the end of this challenge, you’ll be equipped with the tools to recognize and analyze different AI models, helping you optimize interactions and improve security defenses.

​
Bear 4: Prompt Injection

In this final challenge, you’ll learn prompt injection techniques to manipulate the output of a large language model (LLM).

​
What You’ll Learn

  • How to work around constraints and limitations imposed on the model’s behavior.
  • Techniques for cleverly using prompt hacking to get the model to reveal restricted information.
This challenge will test your ability to exploit LLMs and will teach you how to craft queries that bypass restrictions.

​
Congratulations! You’ve Completed Onboarding.

You’ve just completed the first step in your journey with Crucible! You now have the foundational knowledge to get going and discover more features. Here’s what you can do next:
  1. Join the community:
    • Engage with other users and get tips on how to optimize your use of the Dreadnode Platform. Join our Discord to ask questions and share experiences.
  2. Complete more challenges:
    • Ready for more? Explore the rest of our Crucible challenges to continue sharpening your AI/ML hacking skills.
  3. Support and resources: