A company has a machine learning model that is a critical business asset. They want to protect the model's confidentiality while still allowing it to be served to customers via an API. Which approach best balances these requirements?

Use model watermarking to track the model's usage

Deploy the model in a secure enclave (e.g., confidential computing) and expose only the API

Limit the number of API requests per customer

Encrypt the model file and provide the decryption key to customers

A retail company is building a recommendation model using customer purchase history. They want to use the data for training but need to reduce the risk of re-identifying individual customers. Which technique would best achieve this?

Store the data in a secure data lake with access controls

Pseudonymize the data by replacing customer names with unique identifiers

Delete all purchase history data after one year

Encrypt the data with a strong encryption algorithm

A cybersecurity firm has developed an intrusion detection model. They are concerned about adversarial attacks where an attacker crafts inputs to evade detection. Which approach would best improve the model's resilience to such attacks?

Train the model on a dataset that includes adversarial examples

Use a more complex model architecture

Encrypt the model's input data before feeding it to the model

Limit the number of features used by the model

A government agency is building a model to predict traffic patterns using location data from citizens. They must ensure that the data is protected from unauthorized access and that the data is not altered during storage. Which combination of controls should they implement?

Encrypt the data at rest and use digital signatures to verify data integrity

Store the data in a private cloud and use a VPN for access

Anonymize the data and use differential privacy for queries

Use role-based access control (RBAC) and audit logs

A company has deployed a machine learning model for image recognition. They are concerned about adversarial attacks that could cause the model to misclassify images. They want to implement a defense that is effective and does not significantly degrade model performance. Which approach is most appropriate?

Restrict the model's API to only accept images from trusted sources

Use adversarial training by augmenting the training set with adversarial examples

Add a preprocessing step that applies a median filter to all input images

Use a more complex model with more layers

Certified Cloud Security Professional

Domain 2Objective 9

Comprehend Data Protection of Artificial Intelligence (AI) and Machine Learning (ML) Data CCSP Practice Questions (Page 4)

Part of the Cloud Data Security domain, which accounts for 20% of the CCSP exam. ISC2 does not publish an official question count, but from its 180-minute exam (~70–120 total, ~14–24 in this domain), expect 2–3 from this objective — we provide 24 practice questions to prepare you well beyond it. (estimate)

24questions here

5free pages

4concepts

20%of the exam

Questions 16–20

16expert · hard
A company has a machine learning model that is a critical business asset. They want to protect the model's confidentiality while still allowing it to be served to customers via an API. Which approach best balances these requirements?
Select an answer first
17application · medium
A retail company is building a recommendation model using customer purchase history. They want to use the data for training but need to reduce the risk of re-identifying individual customers. Which technique would best achieve this?
Select an answer first
18expert · hard
A cybersecurity firm has developed an intrusion detection model. They are concerned about adversarial attacks where an attacker crafts inputs to evade detection. Which approach would best improve the model's resilience to such attacks?
Select an answer first
19application · medium
A government agency is building a model to predict traffic patterns using location data from citizens. They must ensure that the data is protected from unauthorized access and that the data is not altered during storage. Which combination of controls should they implement?
Select an answer first
20expert · hard
A company has deployed a machine learning model for image recognition. They are concerned about adversarial attacks that could cause the model to misclassify images. They want to implement a defense that is effective and does not significantly degrade model performance. Which approach is most appropriate?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to CCSP

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by ISC2. “CCSP” is a trademark of its owner, used for identification only.