LLM Security Quiz: Prompt Injection & Data Leaks

1. Which of the following is NOT a common LLM security risk?

Excessive server disk space usage

Prompt injection attacks

Model hallucinations leaking sensitive data

Training data extraction

2. What is the primary goal of Prompt Security & LLM Risks (SecAI+) frameworks?

Establish standards for identifying and mitigating LLM security threats

Increase LLM processing speed

Reduce the cost of deploying LLM applications

Simplify natural language understanding

3. Monitoring and logging LLM interactions is essential for detecting security incidents.

True

False

4. Which best describes the relationship between LLM transparency and security?

Greater transparency helps identify vulnerabilities but may expose attack vectors

Transparency and security are unrelated

Complete transparency eliminates all security risks

Security requires total opacity of model operations

Submit

6. Context window attacks exploit the LLM's ability to process large amounts of text.

True

False

7. Which defense mechanism limits the types of functions an LLM can execute?

Function whitelisting

Rate limiting

Data encryption

Token caching

8. Protecting system prompts from exposure is critical to maintaining LLM security controls.

True

False

9. What is a system prompt in LLM security context?

Initial instructions defining the model's behavior and constraints

A prompt that executes system commands

A backup copy of user prompts

An automated request to update the model

10. Red teaming LLMs involves security professionals attempting to break or compromise the model.

True

False

11. What is a prompt injection attack?

An attempt to manipulate LLM behavior by inserting malicious instructions into user input

A method to improve LLM accuracy through data injection

A process of training LLMs on large datasets

A technique to reduce model latency

12. Jailbreaking an LLM means bypassing its safety guidelines to generate harmful content.

True

False

13. Which attack extracts sensitive information by asking the LLM targeted questions?

Extraction attack

Denial of service

Man-in-the-middle

Brute force attack

Submit

15. What is membership inference in LLM security?

Determining whether specific data was in the training set

Verifying user authentication credentials

Checking model permissions on a server

Validating API key authenticity

16. Model poisoning refers to corrupting training data to degrade LLM performance.

True

False

17. Which mitigation strategy helps defend against prompt injection?

Input validation and prompt sandboxing

Removing all user inputs from the system

Running the LLM offline only

Disabling all API access

18. What is an indirect prompt injection attack?

Embedding malicious prompts in external data sources retrieved by the LLM

Attacking the LLM model file directly

Compromising the server infrastructure

Flooding the API with legitimate requests

19. A direct prompt injection attack targets user-controlled input directly to the LLM.

True

False

20. Which of the following is a common data leakage risk in LLMs?

The model reproducing sensitive training data in its responses

Users downloading model weights without authorization

Network bandwidth overload during inference

Increased computational cost per query

LLM Security Quiz: Can You Master Prompt Injection & Data Leaks?

1. Which of the following is NOT a common LLM security risk?

2.

What first name or nickname would you like us to use?

2. What is the primary goal of Prompt Security & LLM Risks (SecAI+) frameworks?

3. Monitoring and logging LLM interactions is essential for detecting security incidents.

4. Which best describes the relationship between LLM transparency and security?

5. A____is a technique that adds intentionally misleading information to confuse prompt injection attempts.

6. Context window attacks exploit the LLM's ability to process large amounts of text.

7. Which defense mechanism limits the types of functions an LLM can execute?

8. Protecting system prompts from exposure is critical to maintaining LLM security controls.

9. What is a system prompt in LLM security context?

10. Red teaming LLMs involves security professionals attempting to break or compromise the model.

11. What is a prompt injection attack?

12. Jailbreaking an LLM means bypassing its safety guidelines to generate harmful content.

13. Which attack extracts sensitive information by asking the LLM targeted questions?

14. Differential privacy can help prevent data leakage by adding noise to____.

15. What is membership inference in LLM security?

16. Model poisoning refers to corrupting training data to degrade LLM performance.

17. Which mitigation strategy helps defend against prompt injection?

18. What is an indirect prompt injection attack?

19. A direct prompt injection attack targets user-controlled input directly to the LLM.

20. Which of the following is a common data leakage risk in LLMs?