Security Controls and Monitoring AAISM Practice Questions (Page 1)

Questions 1–5

1. 1application · medium

   A security analyst is reviewing logs from an AI-based facial recognition system. The analyst notices that the system is incorrectly matching individuals to different identities. The analyst suspects an evasion attack. What evidence would support this suspicion?

   • AThe system's training data was found to contain incorrect labels.
   • BThe misclassifications occur only when specific patterns are present in the input images.
   • CThe system's accuracy has been declining gradually over time.
   • DThe misclassifications occur randomly across all users.
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   An organization's AI system processes personal data of EU citizens. The organization must ensure that its AI security controls meet GDPR requirements. During an audit, the auditor asks how the organization ensures that AI decisions are explainable. What is the most appropriate response?

   • AState that the AI model is proprietary and cannot be explained.
   • BUse a black-box model that has been approved by the data protection officer.
   • CImplement explainability tools that provide human-readable justifications for AI decisions.
   • DOnly provide explanations if the individual files a formal complaint.
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   A company's AI model for customer churn prediction is showing a significant drop in performance. The data science team suspects that the model is experiencing concept drift because customer behavior has changed. What is the most appropriate action to confirm concept drift?

   • ACheck the model's training data for signs of data poisoning.
   • BIncrease the model's complexity to better fit the new data.
   • CRetrain the model with the most recent data and see if performance improves.
   • DCompare the model's predictions against actual outcomes over a recent time period.
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A security analyst notices that an AI chatbot is generating offensive and biased responses to certain user inputs. The analyst suspects that the model has been attacked. What type of attack is most likely?

   • AData poisoning
   • BPrompt injection
   • CModel inversion
   • DEvasion attack
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A security team is investigating a potential data poisoning attack on a machine learning model used for credit scoring. The team has identified that the model's performance has degraded over time. They need to determine if the degradation is due to data poisoning or natural model drift. What is the most effective way to distinguish between the two?

   • ACheck the model's accuracy on a validation set from the original training period.
   • BMonitor the model's output distribution over time.
   • CRetrain the model with new data and see if performance improves.
   • DAnalyze the training data for anomalies and compare the model's behavior on known benign inputs.
   Reveal Solution Discussion Select an answer first