AI Vulnerability Scanning and Fuzzing COASP Practice Questions (Page 10)

Questions 46–50

1. 46foundation · easy

   How does AI fuzzing differ from conventional fuzzing?

   • AAI fuzzing focuses on causing misclassifications and other model-specific failures, not just crashes
   • BAI fuzzing does not require any input generation
   • CAI fuzzing is only used for image recognition models
   • DAI fuzzing only tests the model's API endpoints
   Reveal Solution Discussion Select an answer first
2. 47application · medium

   A vulnerability scan of a recommendation system reveals that an attacker can perform a membership inference attack to determine whether a specific user's data was used in training. The system's privacy policy requires protecting user data. Which mitigation is most directly effective against membership inference?

   • AApply differential privacy during model training.
   • BUse a larger training dataset.
   • CAdd rate limiting to the inference endpoint.
   • DEncrypt the model file at rest.
   Reveal Solution Discussion Select an answer first
4. 48application · medium

   A fuzzing campaign on a credit scoring model produces a set of inputs that cause the model to give a high credit score to applicants with very low income and high debt. The team needs to interpret these results. What is the most appropriate conclusion?

   • AThe model is working as intended because the fuzzing inputs are unrealistic.
   • BThe fuzzer is buggy because it generated inputs that the model was not trained on.
   • CThis is a potential model fairness and robustness vulnerability that could be exploited to obtain undeserved credit; it should be reported and mitigated.
   • DThe model is overfitting to the training data, so the team should add more regularization.
   Reveal Solution Discussion Select an answer first
5. 49foundation · easy

   To detect vulnerabilities in an AI model's inference endpoint, which scanning method is most suitable?

   • ASending crafted adversarial inputs to the endpoint and observing the model's outputs
   • BReviewing the model's training logs for unusual loss values
   • CInspecting the physical security of the data center
   • DChecking the model's license compliance
   Reveal Solution Discussion Select an answer first
6. 50foundation · easy

   When interpreting results from an AI vulnerability scan, what should you prioritize?

   • AThe number of false positives
   • BThe total number of test cases run
   • CThe severity and exploitability of the identified weaknesses
   • DThe model's accuracy on the test set
   Reveal Solution Discussion Select an answer first