Bypassing Defensive Controls GOAA Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   What is a typical weakness in anomaly detection systems that can be exploited to bypass them?

   • AThey rely on human review, which is easily fooled
   • BThey are too slow to process real-time traffic
   • CThey always produce false positives, making them unreliable
   • DThey are trained only on normal data, so novel but malicious patterns may not be flagged
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   An organization deploys an AI-based intrusion detection system (IDS) that uses a neural network to detect network anomalies. A red team discovers that the IDS can be evaded by adding small, carefully crafted perturbations to network traffic. The organization wants to evaluate mitigations. Which mitigation is most likely to improve robustness against such adversarial perturbations?

   • ADeploying the model on more powerful hardware to process traffic faster.
   • BAdversarial training by augmenting the training dataset with adversarial examples generated during training.
   • CIncreasing the model's complexity by adding more layers and neurons to improve accuracy.
   • DReducing the model's input feature space by removing features that are not frequently used.
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   Which of the following is a common weakness in input sanitization that an attacker might exploit to bypass an AI-based defensive control?

   • AThe sanitization process uses a whitelist of allowed characters, which is always effective
   • BThe sanitization process uses machine learning to detect anomalies, which cannot be bypassed
   • CThe sanitization process is applied after the AI model makes its decision
   • DThe sanitization process only strips obvious malicious patterns, allowing adversarial perturbations to pass through
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A security team has deployed an AI-based malware detector that uses both static and dynamic analysis. The detector is robust to adversarial examples in static analysis but vulnerable in dynamic analysis. The team wants to improve overall security without significantly increasing false positives. Which countermeasure is most appropriate?

   • ARemove dynamic analysis and rely solely on static analysis.
   • BIncrease the sensitivity of the static analysis model to catch more malware.
   • CUse a simple rule-based system for dynamic analysis instead of a deep learning model.
   • DApply adversarial training to the dynamic analysis model using known evasion techniques.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   After a red-team exercise, a security analyst discovers that an image-classification model used for content moderation can be evaded by adding a small watermark-like pattern. The team wants to propose a mitigation that is least likely to degrade the model's performance on legitimate images. Which mitigation should they recommend?

   • AAdversarial training by augmenting the training set with the discovered watermark pattern and similar perturbations.
   • BInput sanitization by removing all watermarks from every image before classification.
   • CIncrease the model's confidence threshold for all classifications.
   • DReplace the model with a simpler linear classifier that is less susceptible to adversarial examples.
   Reveal Solution Discussion Select an answer first