The answer is adversarial training and gradient masking, such as defensive distillation. Adversarial training works by injecting perturbed examples into the training set, forcing the model to learn robust decision boundaries that resist small, malicious input changes. Gradient masking, often implemented through defensive distillation, hides the model’s gradient information that attackers rely on to craft evasion attacks, making it significantly harder to generate effective perturbations. On the CompTIA AI+ AI0-001 exam, this question tests your understanding of active versus passive defenses against adversarial evasion attacks on image classifiers. A common trap is confusing data augmentation, which improves generalization but does not specifically counter adversarial perturbations, with adversarial training. Another pitfall is overlooking gradient masking as a distinct technique—many candidates remember only adversarial training. For a memory tip, think of “train and mask”: adversarial training hardens the model’s surface, while gradient masking hides the map attackers use to find weaknesses.
AI0-001 AI Security, Ethics and Governance Practice Question
This AI0-001 practice question tests your understanding of ai security, ethics and governance. Read the scenario carefully and evaluate each option against the stated constraints before committing to an answer. After answering, compare your reasoning against the explanation and wrong-answer breakdown below. Once you have made your selection, read the full explanation to reinforce the concept and understand why each distractor is designed to mislead on exam day.
Which TWO of the following are effective defenses against adversarial evasion attacks on image classifiers?
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Gradient masking
Adversarial training and gradient masking (e.g., defensive distillation) are common defenses. Data augmentation helps generalization but not specifically against adversarial perturbations; feature squeezing reduces input complexity; input validation is generic.
Key principle: Count usable hosts — not total addresses — and remember that the network and broadcast addresses are not available to hosts in standard IPv4 subnets.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
✗
Data augmentation
Why it's wrong here
Data augmentation adds noise/transformations but not targeted adversarial perturbations.
✓
Gradient masking
Why this is correct
Gradient masking obscures gradient information to prevent crafting adversarial examples.
Input validation checks for malicious inputs but adversarial perturbations are often undetectable.
✗
Feature squeezing
Why it's wrong here
Feature squeezing reduces input precision to remove adversarial perturbations, but it is not always effective.
Common exam traps
Common exam trap: usable hosts are not the same as total addresses
Subnetting questions often tempt you into counting all addresses. In normal IPv4 subnets, the network and broadcast addresses are not usable host addresses.
Detailed technical explanation
How to think about this question
Subnetting questions test whether you can identify the network, broadcast address, usable range, mask and correct subnet. Slow down enough to calculate the block size correctly.
KKey Concepts to Remember
CIDR notation defines the prefix length.
Block size helps identify subnet boundaries.
Network and broadcast addresses are not usable hosts in normal IPv4 subnets.
The required host count determines the smallest suitable subnet.
TExam Day Tips
→Write the block size before choosing the subnet.
→Check whether the question asks for hosts, subnets or a specific address range.
→Do not confuse /24, /25, /26 and /27 host counts.
Key takeaway
Count usable hosts — not total addresses — and remember that the network and broadcast addresses are not available to hosts in standard IPv4 subnets.
Real-world example
How this comes up in practice
A network engineer segments a warehouse floor into three subnets: 20 scanners, 5 printers, and 2 management hosts. Picking the wrong mask wastes addresses or leaves too few usable hosts. Exam questions test whether you can apply CIDR notation, calculate block size, and identify the correct usable-host range for a given prefix.
What to study next
Got this wrong? Here's your next step.
Review block sizes, usable host formulas (2^n − 2), and how to find network and broadcast addresses for /24 through /30. Then practise related AI0-001 subnetting questions on CIDR, address ranges, and subnet selection.
AI Security, Ethics and Governance — This question tests AI Security, Ethics and Governance — CIDR notation defines the prefix length..
What is the correct answer to this question?
The correct answer is: Gradient masking — Adversarial training and gradient masking (e.g., defensive distillation) are common defenses. Data augmentation helps generalization but not specifically against adversarial perturbations; feature squeezing reduces input complexity; input validation is generic.
What should I do if I get this AI0-001 question wrong?
Review block sizes, usable host formulas (2^n − 2), and how to find network and broadcast addresses for /24 through /30. Then practise related AI0-001 subnetting questions on CIDR, address ranges, and subnet selection.
What is the key concept behind this question?
CIDR notation defines the prefix length.
About these practice questions
Courseiva creates original exam-style practice questions with explanations and wrong-answer analysis. It does not publish real exam questions, exam dumps, or protected exam content. Learn why practice questions differ from exam dumps →
These questions test the same concept from different angles. Work through them to make sure you can recognise it however the exam phrases it.
Variation 1. Which TWO of the following are best practices for securing an AI model against adversarial attacks?
easy
A.Model pruning to reduce the number of parameters.
✓ B.Adversarial training with perturbed examples.
✓ C.Input sanitization and validation.
D.Increasing model complexity to capture more patterns.
E.Hyperparameter optimization using grid search.
Why B: Option B is correct because adversarial training explicitly augments the training dataset with perturbed examples (e.g., using FGSM or PGD attacks) to teach the model to recognize and resist malicious inputs. This method directly hardens the model against evasion attacks by improving its decision boundary robustness.
Variation 2. Which TWO of the following are effective defenses against adversarial examples in AI systems?
medium
✓ A.Train the model with adversarial examples (adversarial training)
B.Use an ensemble of models and majority voting
C.Increase the model's sensitivity to input changes
✓ D.Implement input sanitization and feature squeezing
E.Reduce model complexity through pruning
Why A: Options B and D are correct. Adversarial training and input sanitization/denoising are defenses. Option A is wrong because model compression may remove defenses. Option C is wrong because increasing model sensitivity can make it more vulnerable. Option E is wrong because ensemble voting is not a direct defense against adversarial examples.
Variation 3. A healthcare AI system misdiagnosed patients due to adversarial inputs. What security measure should be prioritized?
medium
A.Encrypt all patient data
B.Use stronger authentication
C.Regular software updates
✓ D.Implement adversarial training
Why D: Option C (Implement adversarial training) is correct because adversarial training makes the model robust to input manipulation. Option A (Encrypt all patient data) protects data privacy but not model integrity. Option B (Use stronger authentication) is for access control. Option D (Regular software updates) is general maintenance.
Last reviewed: Jun 23, 2026
Question Discussion
Share a tip, memory trick, or ask about the reasoning behind this question. Do not post real exam questions, leaked content, braindumps, or copyrighted exam material. Comments are moderated and may be removed without notice.
This AI0-001 practice question is part of Courseiva's free CompTIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI0-001 exam.
Question Discussion
Share a tip, memory trick, or ask about the reasoning behind this question. Do not post real exam questions, leaked content, braindumps, or copyrighted exam material. Comments are moderated and may be removed without notice.
Sign in to join the discussion.