Creating Malicious Software with AI GOAA Practice Questions (Page 6)

Questions 26–30

1. 26foundation · easy

   Which technique is commonly used to craft adversarial examples that evade AI-based malware detection?

   • ARecompiling the malware with a different compiler to change its binary structure.
   • BGradient-based optimization to find small perturbations that cause misclassification.
   • CPolymorphic code generation that changes the malware's signature each time it runs.
   • DUsing a packer to compress the malware and hide its true code.
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A security vendor is developing an AI-based malware detector and wants to test its robustness against adversarial examples. The vendor has a large dataset of benign and malicious files. Which approach is most effective for evaluating robustness?

   • ATest the detector on a separate hold-out set of benign and malicious files that were not used in training.
   • BUse a third-party penetration testing service to attempt to bypass the detector.
   • CGenerate adversarial examples from the dataset using multiple attack algorithms and measure the detector's accuracy on them.
   • DMeasure the detector's false positive rate on benign files.
   Reveal Solution Discussion Select an answer first
4. 28foundation · easy

   Which statement best describes how an AI model can be used to generate malicious code?

   • AThe model automatically patches vulnerabilities in the target system to allow malware delivery.
   • BThe model compiles source code into machine code, making it harder for antivirus to inspect.
   • CThe model encrypts the malware payload so that it cannot be detected by signature-based scanners.
   • DThe model learns patterns from existing malware and produces new, unseen code variants that may retain malicious functionality.
   Reveal Solution Discussion Select an answer first
5. 29foundation · easy

   What is the primary goal of an adversarial example in the context of evading AI-based malware detection?

   • ATo encrypt the malware payload with a strong algorithm.
   • BTo add small, carefully crafted perturbations to malware so that the detection model misclassifies it as benign.
   • CTo remove all network indicators from the malware to avoid network-based detection.
   • DTo increase the malware's execution speed so it completes before detection.
   Reveal Solution Discussion Select an answer first
6. 30expert · hard

   A deepfake detection company is developing a product that must detect both image and video deepfakes. They have a large dataset of real images and videos, but limited deepfake samples. They decide to use a GAN to generate additional deepfake samples for training. What is the most significant risk of this approach?

   • AThe GAN may introduce bias in the training data, causing the detector to be biased against certain demographics.
   • BThe GAN-generated deepfakes may not be representative of real-world deepfakes, leading to poor generalization.
   • CThe GAN may require too much computational resources to train.
   • DThe GAN may produce low-quality images that are easy to detect, making the training data too easy.
   Reveal Solution Discussion Select an answer first