2.9 End of Chapter Activities
Exercises
Discussion Questions
- Why is the AI arms race particularly relevant in ML security?
- How can organizations stay ahead of attackers in the AI arms race?
Knowledge Check
Quiz Text Description
1. MultiChoice Activity
Which of the following is a common category of ML attacks?
- Integrity Attacks
- All of the Above
- Privacy Attacks
- Evasion Attacks
2. MultiChoice Activity
In a causative attack, what is the main objective of the attacker?
- To manipulate the training data to degrade classifier accuracy
- To explore the decision boundary of the model
- To manipulate test data for incorrect classification
- To modify the model’s decision boundary
3. MultiChoice Activity
Which type of attack aims to prevent legitimate users from accessing machine learning services?
- Integrity Attacks
- Availability Attacks
- Evasion Attacks
- Privacy Attacks
4. MultiChoice Activity
What is the main difference between black-box and white-box attacks?
- White-box attackers have full access to model architecture, while black-box attackers have limited access
- White-box attacks are more likely to succeed than black-box attacks
- Black-box attacks target model predictions, while white-box attacks target data
- Black-box attackers have full access to model architecture, while white-box attackers have limited access
5. MultiChoice Activity
Which of the following best describes the concept of “security by design”?
- Implementing security features after attacks have occurred
- Hiding system details from attackers to avoid exploitation
- Designing systems with inherent security features from the outset
- Relying on human oversight to ensure security post-deployment
Correct Answers:
- b. All of the Above
- a. To manipulate the training data to degrade classifier accuracy
- b. Availability Attacks
- a. White-box attackers have full access to model architecture, while black-box attackers have limited access
- c. Designing systems with inherent security features from the outset
High Flyer. (2025). Deep Seek. [Large language model]. https://www.deepseek.com/
Prompt: Can you provide end-of-chapter questions for the content? Reviewed and edited by the author.
