AAISM: Advanced in AI Security Management practice questions, page 12 of 19, with answer review and explanations.
Question 56
Personal data used to train AI systems can BEST be protected by:
anonymizing personal data.
hashing personal data.
erasing personal data after training.
ensuring the quality of personal data.
Question 57
Embedding unique identifiers into AI models would BEST help with:
preventing unauthorized access.
detecting adversarial attacks.
tracking ownership.
eliminating AI system biases.
Question 58
Which of the following AI-driven systems should have the MOST stringent recovery time objective (RTO)?
Health support system
Credit risk modeling system
Car navigation system
Industrial control system
Question 59
An organization has requested a developer to apply AI algorithms to existing modules in order to improve customer service quality. At this stage, which of the following should be considered FIRST?
IT management may need to revise the service agreement if AI behavior cannot be predefined.
The organization may need to explain the performance of the applied AI algorithm.
Project sponsors may need to agree on a phased approach in order to ensure safe release.
The developer may need to be held accountable for business inquiries raised by customers.
Question 60
Which of the following BEST describes how supervised learning models help reduce false positives in cybersecurity threat detection?
They dynamically generate new labeled data sets.
They analyze patterns in data to group legitimate activity from actual threats.
They learn from historical labeled data.
They use real-time feature engineering to automatically adjust decision boundaries.