What is data poisoning in the context of AI security?

Stealing the model's parameters via network sniffing

Deleting the model's training logs

Injecting malicious or incorrect data into the training set to corrupt the model

Encrypting the training data to prevent access

A healthcare organization is developing a machine learning model to predict patient readmission risk. They have a limited dataset and are concerned about both data poisoning and adversarial examples. Which defense strategy is most appropriate given the limited data?

Use data augmentation and adversarial training with synthetic samples

Use a complex deep learning model to capture all patterns

Disable all model explainability features

Rely solely on input sanitization at inference time

A security team wants to deploy an AI system that detects phishing emails. They have a large dataset of labeled emails and need the system to automatically learn patterns that distinguish phishing from legitimate messages. Which approach is most appropriate?

Use a deep learning model without any training data

Use a supervised machine learning model trained on the labeled dataset

Use an unsupervised clustering algorithm to group emails without labels

Use a rule-based expert system with manually defined patterns

A security analyst is investigating a potential VM escape vulnerability in a hypervisor. Which condition would increase the likelihood of a successful VM escape attack?

The hypervisor is configured with strict isolation settings

The VM is configured with minimal virtual hardware

The VM has access to a passthrough device, such as a GPU

The hypervisor is running the latest security patches

GIAC Security Essentials

Domain 6Objective 3

Virtualization, Cloud Security, and AI Essentials GSEC Practice Questions (Page 9)

Part of the Data Protection and Emerging Technologies domain, which makes up ~12% of our current practice bank. GIAC (SANS) does not publish an official question count, but from its 240-minute exam (~95–160 total, ~11–19 in this domain), expect 4–6 from this objective — we provide 53 practice questions to prepare you well beyond it. (estimate)

53questions here

11free pages

10concepts

Questions 41–45

41foundation · easy
What is data poisoning in the context of AI security?
Select an answer first
42foundation · easy
Which cloud deployment model is exclusively used by a single organization and may be hosted on-premises or by a third party?
Select an answer first
43expert · hard
A healthcare organization is developing a machine learning model to predict patient readmission risk. They have a limited dataset and are concerned about both data poisoning and adversarial examples. Which defense strategy is most appropriate given the limited data?
Select an answer first
44application · medium
A security team wants to deploy an AI system that detects phishing emails. They have a large dataset of labeled emails and need the system to automatically learn patterns that distinguish phishing from legitimate messages. Which approach is most appropriate?
Select an answer first
45application · medium
A security analyst is investigating a potential VM escape vulnerability in a hypervisor. Which condition would increase the likelihood of a successful VM escape attack?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to GSEC

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by GIAC (SANS). “GSEC” is a trademark of its owner, used for identification only.