CompTIA SecAI+ 1.0 Basic AI Concepts Related to Cybersecurity Practice Questions

Question 1

Which of the following AI model training techniques involves training an algorithm across multiple decentralized edge devices or servers holding local data samples, without exchanging them?

Accepted Answer

Federated learning

Answer

Reinforcement learning

Answer

Supervised learning

Answer

Transfer learning

Question 2

A security analyst is using an AI tool to generate code snippets. To minimize the risk of the model hallucinating or providing insecure code, the analyst provides the model with three specific examples of secure coding patterns before asking for the final output. Which prompt engineering technique is this?

Accepted Answer

Multi-shot (Few-shot) prompting

Answer

Zero-shot prompting

Answer

One-shot prompting

Answer

Chain-of-thought prompting

Question 3

Which component of the Retrieval-Augmented Generation (RAG) architecture is responsible for converting textual data into numerical representations that capture semantic meaning?

Accepted Answer

Embeddings

Answer

Vector storage

Answer

Hyperparameters

Answer

System prompt

Question 4

An organization wants to detect if their proprietary images are being used to train unauthorized external AI models. Which technique should they implement?

Accepted Answer

Watermarking

Answer

Data sanitization

Answer

Tokenization

Answer

Pruning

Question 5

During the AI life cycle, which phase involves 'Human-in-the-loop' to ensure the model aligns with human values and safety guidelines before full automation?

Accepted Answer

Model evaluation/Validation

Answer

Data collection

Answer

Deployment

Answer

Decommissioning

Question 6

Which tool is commonly used to package AI applications and their dependencies to ensure consistent deployment across environments?

Accepted Answer

Virtual Containers (e.g., Docker)

Answer

Spreadsheets

Answer

Hypervisors only

Answer

VPNs

Question 7

Which metric is critical to monitor to detect 'Model Drift'?

Accepted Answer

Model accuracy/performance over time

Answer

Server CPU temperature

Answer

Network bandwidth usage

Answer

Number of registered users

Question 8

To prevent 'Hallucinations' in an AI customer support bot, a company implements a system that grounds the AI's answers strictly in their own knowledge base. This is an implementation of:

Accepted Answer

RAG (Retrieval-Augmented Generation)

Answer

GAN (Generative Adversarial Network)

Answer

Unsupervised Learning

Answer

Data Poisoning

Question 9

Which validation technique involves removing connections in a neural network to reduce model size and complexity without significantly affecting accuracy?

Accepted Answer

Pruning

Answer

Augmentation

Answer

Encryption

Answer

Tokenization

Question 10

Which of the following describes 'Quantization' in AI models?

Accepted Answer

Reducing the precision of the model's parameters (e.g., from 32-bit float to 8-bit integer) to save memory

Answer

Increasing the number of parameters to billions

Answer

Encrypting the model using quantum cryptography

Answer

Training the model on quantum computers

Question 11

Which type of AI model is specifically designed to generate new content (text, images, audio) based on learned patterns?

Accepted Answer

Generative AI

Answer

Discriminative AI

Answer

Linear Regression

Answer

K-Means Clustering

Question 12

Which control helps ensure that an AI model is not trained on sensitive PII data inadvertently?

Accepted Answer

Data Cleansing / Scrubbing

Answer

Data Augmentation

Answer

Increasing Learning Rate

Answer

Model Pruning

Question 13

Which of the following is considered a 'Small Language Model' (SLM) typically designed for edge deployment?

Accepted Answer

Microsoft Phi-3 / Google Gemma

Answer

GPT-4

Answer

Claude 3 Opus

Answer

Gemini Ultra

Question 14

What is model drift?

Accepted Answer

A change in model performance as data patterns change over time

Answer

A network cable moving between ports

Answer

A disk cloning method

Answer

A password reset workflow

Question 15

Which metric focuses on the proportion of predicted positive results that were actually positive?

Accepted Answer

Precision

Answer

Latency only

Answer

Availability

Answer

Entropy

Question 16

Which metric focuses on how many actual positive cases the model found?

Accepted Answer

Recall

Answer

Packet loss

Answer

RPO

Answer

Uptime

Question 17

Which approach keeps raw training data on local devices while sharing model updates?

Accepted Answer

Federated learning

Answer

Static routing

Answer

Full disk imaging

Answer

Screen scraping

Question 18

What is hallucination in generative AI?

Accepted Answer

A plausible but false or unsupported output

Answer

A guaranteed encrypted response

Answer

A firewall drop rule

Answer

A disk sector error

Question 19

Which dataset split is used to tune model choices during development without using the final test set?

Accepted Answer

Validation set

Answer

Production logs only

Answer

Backup set

Answer

DNS cache

Question 20

Which technique retrieves trusted documents and provides them to an LLM at answer time?

Accepted Answer

Retrieval-augmented generation

Answer

Disk mirroring

Answer

NAT traversal

Answer

Static routing

Question 21

Which concept describes limiting generated answers to approved source material?

Accepted Answer

Grounding

Answer

Overclocking

Answer

Subnetting

Answer

Fragmentation

Question 22

Which evaluation dataset should remain unused during training and tuning until final assessment?

Accepted Answer

Test set

Answer

Training set

Answer

Temporary cache

Answer

DNS zone

CompTIA SecAI+ Practice Questions: 1.0 Basic AI Concepts Related to Cybersecurity

Q1. Which of the following AI model training techniques involves training an algorithm across multiple decentralized edge devices or servers holding local data samples, without exchanging them?

Q3. Which component of the Retrieval-Augmented Generation (RAG) architecture is responsible for converting textual data into numerical representations that capture semantic meaning?

Q4. An organization wants to detect if their proprietary images are being used to train unauthorized external AI models. Which technique should they implement?

Q5. During the AI life cycle, which phase involves 'Human-in-the-loop' to ensure the model aligns with human values and safety guidelines before full automation?

Q6. Which tool is commonly used to package AI applications and their dependencies to ensure consistent deployment across environments?

Q7. Which metric is critical to monitor to detect 'Model Drift'?

Q8. To prevent 'Hallucinations' in an AI customer support bot, a company implements a system that grounds the AI's answers strictly in their own knowledge base. This is an implementation of:

Q9. Which validation technique involves removing connections in a neural network to reduce model size and complexity without significantly affecting accuracy?

Q10. Which of the following describes 'Quantization' in AI models?

Q11. Which type of AI model is specifically designed to generate new content (text, images, audio) based on learned patterns?

Q12. Which control helps ensure that an AI model is not trained on sensitive PII data inadvertently?

Q13. Which of the following is considered a 'Small Language Model' (SLM) typically designed for edge deployment?

Q14. What is model drift?

Q15. Which metric focuses on the proportion of predicted positive results that were actually positive?

Q16. Which metric focuses on how many actual positive cases the model found?

Q17. Which approach keeps raw training data on local devices while sharing model updates?

Q18. What is hallucination in generative AI?

Q19. Which dataset split is used to tune model choices during development without using the final test set?

Q20. Which technique retrieves trusted documents and provides them to an LLM at answer time?

Q21. Which concept describes limiting generated answers to approved source material?

Q22. Which evaluation dataset should remain unused during training and tuning until final assessment?

More CompTIA SecAI+ practice topics