A manufacturing company uses a machine learning model to predict equipment failures. After a routine update, the model's accuracy drops significantly on certain sensor inputs, and the operations team suspects the training data was tampered with. Which action would best confirm this suspicion?

Review the model's training logs for any anomalies

Test the model against a small set of known-good sensor inputs

Compare the current training dataset's checksums against the original dataset's checksums

Retrain the model on the current dataset and compare accuracy

A healthcare startup is training a diagnostic model on patient records that include diagnosis codes and lab results. They must allow data scientists to develop features without exposing individual patient identities, while still enabling the model to learn correlations. Which approach best meets this requirement?

Apply k-anonymity to the dataset before sharing it with the data science team

Encrypt the dataset with a customer-managed key and share the key with the data science team

Remove all diagnosis codes and lab results from the dataset

Store the dataset in a private cloud with strict access controls

A research institute has trained a medical diagnosis model on sensitive patient data. They are concerned about membership inference attacks, where an attacker can determine whether a specific patient's data was used in training. Which mitigation strategy would be most effective?

Apply differential privacy during model training

Use model watermarking to track the model's distribution

Limit the number of API requests per user

Encrypt the model file with a strong encryption algorithm

Certified Cloud Security Professional

Domain 2Objective 9

Comprehend Data Protection of Artificial Intelligence (AI) and Machine Learning (ML) Data CCSP Practice Questions (Page 5)

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 21–24

21expert · hard
A manufacturing company uses a machine learning model to predict equipment failures. After a routine update, the model's accuracy drops significantly on certain sensor inputs, and the operations team suspects the training data was tampered with. Which action would best confirm this suspicion?
Select an answer first
22application · medium
A healthcare startup is training a diagnostic model on patient records that include diagnosis codes and lab results. They must allow data scientists to develop features without exposing individual patient identities, while still enabling the model to learn correlations. Which approach best meets this requirement?
Select an answer first
23expert · hard
A research institute has trained a medical diagnosis model on sensitive patient data. They are concerned about membership inference attacks, where an attacker can determine whether a specific patient's data was used in training. Which mitigation strategy would be most effective?
Select an answer first
24foundation · easy
Which type of attack involves an attacker injecting malicious samples into the training data to cause the model to misbehave in a specific way?
Select an answer first

Finished these 4 questions?

Review the revealed explanations, or continue through the curriculum.

No more pagesBack 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.

Comprehend Data Protection of Artificial Intelligence (AI) and Machine Learning (ML) Data CCSP Practice Questions (Page 5)

Questions 21–24

A manufacturing company uses a machine learning model to predict equipment failures. After a routine update, the model's accuracy drops significantly on certain sensor inputs, and the operations team suspects the training data was tampered with. Which action would best confirm this suspicion?

A research institute has trained a medical diagnosis model on sensitive patient data. They are concerned about membership inference attacks, where an attacker can determine whether a specific patient's data was used in training. Which mitigation strategy would be most effective?

Which type of attack involves an attacker injecting malicious samples into the training data to cause the model to misbehave in a specific way?

Keep studying — Cloud Data Security