Comprehend Data Protection of Artificial Intelligence (AI) and Machine Learning (ML) Data CCSP Practice Questions (Page 1)

Questions 1–5

1. 1expert · hard

   A company has deployed a machine learning model for fraud detection. They notice that the model's performance degrades over time, and they suspect that an attacker is submitting crafted transactions to influence the model's future updates. Which type of attack is this, and what is the best mitigation?

   • AAdversarial example; use adversarial training
   • BData poisoning; implement robust validation of training data and monitor for anomalies
   • CMembership inference; restrict API access to the model
   • DModel inversion; use differential privacy during training
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   Which privacy technique replaces identifying values in a training data set with artificial identifiers while still allowing records to be linked across tables?

   • AAnonymization
   • BPseudonymization
   • CDifferential privacy
   • DTokenization
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A social media company is building a sentiment analysis model using user posts. They want to protect user privacy while still being able to analyze trends. They decide to use differential privacy when training the model. What is the primary benefit of this approach?

   • AIt prevents the model from being stolen by unauthorized parties
   • BIt encrypts the training data so that only authorized users can access it
   • CIt ensures that the model cannot be used to infer any individual user's data
   • DIt anonymizes the training data by removing all personal identifiers
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   Which technique embeds a unique, imperceptible pattern into a trained model so that unauthorized copies can be traced back to their source?

   • AModel watermarking
   • BModel encryption
   • CDifferential privacy
   • DAdversarial training
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   A technology company has developed a proprietary language model and wants to protect it from being copied or redistributed without authorization. Which technique would best help the company prove ownership if the model is leaked?

   • AEncrypt the model file with a strong encryption algorithm
   • BUse differential privacy during training
   • CLimit access to the model via API only
   • DEmbed a unique watermark in the model's parameters
   Reveal Solution Discussion Select an answer first