Bypassing Defensive Controls GOAA Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   What is the purpose of evaluating the effectiveness of a bypass technique against a defensive control?

   • ATo determine whether the technique successfully evades the control and to inform mitigation strategies
   • BTo replace the defensive control with a more complex one
   • CTo prove that the defensive control is completely secure
   • DTo document the attack for legal purposes only
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   Which of the following is a key characteristic of a realistic red-team scenario for testing AI defenses?

   • AThe red team operates under constraints similar to those of a real attacker, such as limited knowledge and resources
   • BThe red team is given unlimited time and budget to find any possible weakness
   • CThe red team has full access to the AI model's training data and parameters
   • DThe red team only tests the AI model in isolation, without considering the surrounding system
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   Which of the following is a query-based method for evading AI-powered defenses?

   • ATraining a surrogate model on the target model's outputs
   • BExploiting a vulnerability in the model's API to gain shell access
   • CUsing the target model's training data to retrain it with adversarial examples
   • DSubmitting carefully crafted inputs to the target model and observing its outputs to construct an adversarial example
   Reveal Solution Discussion Select an answer first
5. 4application · medium

   A company deploys an AI-based intrusion detection system (IDS) that flags network flows as malicious based on features like packet size and duration. A penetration tester wants to evade detection while maintaining the actual attack's effectiveness. The tester has observed that the IDS is sensitive to packet-size thresholds. Which action is most likely to bypass the IDS while still delivering the attack payload?

   • ASend the attack during off-peak hours to reduce the IDS's alert threshold.
   • BFragment the payload into smaller packets that fall below the IDS's size threshold.
   • CUse a different source IP address for each packet.
   • DEncrypt the entire attack traffic so the IDS cannot inspect it.
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   A security operations center uses a deep learning model to classify malware from executable files. A penetration tester has successfully evaded the model using a gradient-based attack on a local copy of the model. However, when the same adversarial executable is submitted to the production model, it is still detected. Which factor most likely explains this failure?

   • AThe adversarial executable was too large to be processed by the production model.
   • BThe production model uses a different file format than the local copy.
   • CThe production model has different parameters or training data than the local copy, reducing transferability.
   • DThe gradient-based attack was not applied to the correct feature space.
   Reveal Solution Discussion Select an answer first