Wirewalk Instituteof AI
sec-340 Expert Available

Adversarial Machine Learning: Evasion, Poisoning and Model Extraction

For ML engineers and security researchers who need to attack and defend machine learning models directly: evasion, poisoning, backdoors and model extraction, evaluated rigorously.

Level
Expert
Study time
20 hours
Lessons
19 in 6 modules
Exam
25 questions

About this course

Long before chatbots, machine learning models in fraud detection, malware classification and computer vision were being fooled, poisoned and copied. Those attacks still apply, now joined by their generative AI counterparts. This course teaches adversarial machine learning with the vocabulary of NIST AI 100-2 E2025 (attacker goals, knowledge and capabilities across the lifecycle) and maps it to MITRE ATLAS techniques.

You will implement gradient-based evasion (FGSM, PGD, Carlini-Wagner), black-box and transfer attacks, and domain-specific evasion against tabular fraud models and malware classifiers where feature constraints matter. You will study availability and targeted poisoning, backdoor trigger design, and poisoning of web-scale, fine-tuning and instruction data. The extraction module covers stealing models through prediction APIs and distilling LLM behavior, with defenses such as rate limiting, output perturbation and watermarking. The defense module treats adversarial training, certified robustness and randomized smoothing, and above all how to evaluate defenses correctly against adaptive attacks and gradient masking, using the Adversarial Robustness Toolbox and RobustBench. The course closes with prioritizing robustness work, building it into model release, and a capstone on a hypothetical fraud classifier. Privacy attacks are covered in sec-330 and application-level prompt attacks in sec-210. The Wirewalk Institute certificate attests to completion and assessment of this course; it is not endorsed by NIST, MITRE, the maintainers of the Adversarial Robustness Toolbox or RobustBench, or any vendor named here.

What you will be able to do

  • Describe an adversarial threat model using NIST AI 100-2 attacker goals, knowledge and capabilities
  • Implement white-box and black-box evasion attacks and adapt them to constrained domains such as fraud and malware
  • Explain availability poisoning, targeted poisoning and backdoors and estimate their feasibility for a given data pipeline
  • Assess model extraction risk for a prediction or LLM API and select proportionate defenses
  • Apply adversarial training and certified robustness and explain their costs
  • Evaluate a defense correctly using adaptive attacks and detect gradient masking
  • Integrate robustness testing into a model release process

Syllabus

6 modules · 19 lessons · about 20 hours

Module 1: Threat Models for Adversarial ML

The vocabulary and structure for reasoning about attacks on models.
3 lessons
  1. Attacker Goals, Knowledge and Capabilities in NIST AI 100-2
  2. Predictive and Generative AI Attack Classes Compared
  3. Mapping Adversarial ML to MITRE ATLAS Techniques

Module 2: Evasion Attacks

Crafting inputs that cause wrong outputs at inference time.
3 lessons
  1. Gradient-Based Attacks: FGSM, PGD and Carlini-Wagner
  2. Black-Box, Query-Based and Transfer Attacks
  3. Constrained and Physical-World Evasion: Fraud, Malware and Vision Models

Module 3: Poisoning and Backdoors

Corrupting models through the data they learn from.
3 lessons
  1. Availability and Targeted Poisoning
  2. Backdoor Attacks and Trigger Design
  3. Poisoning Web-Scale, Fine-Tuning and Instruction Data

Module 4: Model Extraction

Stealing functionality through the interfaces you expose.
3 lessons
  1. Model Extraction Through Prediction APIs
  2. Distillation and Behavior Stealing from LLM APIs
  3. Defending APIs: Rate Limits, Output Perturbation and Watermarking

Module 5: Defenses and Rigorous Evaluation

What works, what it costs and how to know.
4 lessons
  1. Adversarial Training and Its Costs
  2. Certified Robustness and Randomized Smoothing
  3. Evaluating Defenses Against Adaptive Attacks and Detecting Gradient Masking
  4. Tooling: Adversarial Robustness Toolbox and RobustBench

Module 6: Applied Robustness Program

Prioritizing and operationalizing robustness.
3 lessons
  1. Prioritizing Which Models Need Robustness Work
  2. Robustness Testing in the Model Release Process
  3. Capstone: Attack and Harden a Hypothetical Fraud Classifier

Exam and certificate

The exam unlocks once you have completed every lesson. It draws 25 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 60 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.

Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.

The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by NIST, MITRE ATLAS, Adversarial Robustness Toolbox, RobustBench, or PyTorch. Product and company names are trademarks of their respective owners.

More in AI Security: Securing AI Systems

All 13 courses in this track
sec-001 Free intro Available

Using AI Safely at Work: The Essentials

Follow one fictional US distributor through four everyday AI mistakes (a guarantor's credit file uploaded to a personal account, a customer letter promising a fee waiver that does not exist, a vendor email with hidden instructions for the assistant, and a note-taker that sent a customer the internal margin floor) and practice the decisions that prevent or contain each one in three short Decision Labs.

1 h · 7 lessons Free intro
sec-101 Basic Available

Using AI Safely: Data Handling for Every Employee

A plain-language course on handling data safely with AI tools: what happens to what you type or upload, how to classify before you paste, how connected apps and agents change the picture, and how to report a mistake quickly.

6 h · 15 lessons See price at checkout
sec-201 Intermediate Available

AI Security Fundamentals: The OWASP Top 10 for LLMs

A practitioner course on securing applications that use large language models, organized around the 2026 OWASP Top 10 for LLM Applications. Every risk comes with a hypothetical scenario, the root cause, and concrete defenses in code and configuration.

14 h · 18 lessons See price at checkout