GIAC GOAA Certification Sample Questions
The purpose of this Sample Question Set is to provide you with information about the GIAC Offensive AI Analyst (GOAA) exam. These sample questions will make you very familiar with both the type and the difficulty level of the questions on the GOAA certification test. To get familiar with real exam environment, we suggest you try our Sample GIAC GOAA Certification Practice Exam. This sample practice exam gives you the feeling of reality and is a clue to the questions asked in the actual GIAC Offensive AI Analyst (GOAA) certification exam.
These sample questions are simple and basic questions that represent likeness to the real GIAC Offensive AI Analyst exam questions. To assess your readiness and performance with real-time scenario based questions, we suggest you prepare with our Premium GIAC GOAA Certification Practice Exam. When you solve real time scenario based questions practically, you come across many difficulties that give you an opportunity to improve.
GIAC GOAA Sample Questions:
01. A design note distinguishes a diffusion image model from a GAN.
Which description of a diffusion model is accurate?
a) It produces an image in a single forward pass of a generator that was trained against a discriminator.
b) It retrieves the closest matching real image from a stored dataset and returns that image unchanged.
c) It starts from random noise and iteratively denoises toward a coherent image under the learned model.
d) It converts incoming text into sub-word tokens for a downstream language model to then process.
02. An authorized web-application test has confirmed a command-injection point. The client asks which two defenses to implement to remediate it.
Which two most directly address command injection? (Choose two.)
a) Invoke the target program directly with an argument array instead of passing a string to a shell.
b) Add SPF, DKIM and DMARC authentication records to the organization's own mail domain for protection.
c) Lower the sampling temperature of the LLM that is used during the testing.
d) Allowlist and strictly validate the acceptable input values first.
e) Issue a brand-new TLS certificate for the application's domain.
03. After a successful deepfake-phishing simulation, the client asks for a user-facing recommendation that does not rely on technology to spot fakes.
Which recommendation best fits that request?
a) Asking users to carefully study many example deepfakes so that they can reliably recognize one on sight during a live call later.
b) A verification procedure for sensitive requests (callback, code word, or dual authorization) applied regardless of appearance.
c) Installing a deepfake-detection browser extension on every corporate endpoint in the fleet.
d) Reducing the number of executives who appear in the organization's public media.
04. A tester wants to describe, at a conceptual level, how an AI model assists an evasion workflow against signature-based antivirus without claiming it defeats behavioral detection.
Which description is accurate and appropriately bounded?
a) The model rewrites the payload so it performs entirely different, benign actions yet still achieves the same malicious goal.
b) The model disables the endpoint's behavioral detection engine as a part of the generation step itself.
c) The model guarantees that the resulting variant will never be detected by any control at all.
d) It accelerates producing obfuscated variants, stressing signatures, while runtime behavioral detection is unaffected.
05. A defender maps email controls to the threats they address.
What does DMARC add on top of SPF and DKIM?
a) It publishes an alignment-and-handling policy for mail failing SPF/DKIM, and provides reporting.
b) It encrypts the entire message body end to end between the sending server and the recipient's client.
c) It scans every inbound attachment for malware inside an isolated detonation sandbox before delivery.
d) It outright blocks any inbound email that happens to contain a hyperlink to an external website.
06. An engagement uses an autonomous agent to probe a web application's inputs iteratively.
Which control most directly prevents the agent from acting outside the authorized target?
a) Give the agent broad, unrestricted internet access so that it can be thorough.
b) Increase the model's temperature in order to encourage more creative probing.
c) Disable the request logging in order to reduce overhead.
d) Restrict the agent to an in-scope allowlist of hosts and endpoints, so next-request decisions stay authorized.
07. An assessment notes a technique category called "living off the land" (LOLBins) as an evasion approach.
Why does using built-in, signed system binaries help an attacker evade some controls?
a) System binaries are invisible to all behavioral monitoring on the host purely because they are digitally signed by the vendor.
b) Using built-in system binaries automatically encrypts the attacker's payload.
c) Signed, trusted system utilities raise fewer novel-binary and reputation alerts than an unknown new executable does.
d) Running a system binary disables AMSI for the duration of its execution.
08. An organization wants a forward-looking control so that media it officially produces can later be distinguished from synthetic imitations.
Which approach most directly supports that goal?
a) Content provenance and cryptographic signing (content credentials) on the media the organization publishes.
b) Lowering the resolution of all published media so that deepfakes are harder to base on it.
c) Publishing media far more frequently so that attackers cannot keep up with it.
d) Requiring staff to detect fakes purely by eye before resharing them.
09. A threat briefing claims that AI code-generation "changes the economics" of malware authoring for lower-skilled actors.
Which statement best captures the realistic effect?
a) It removes the need for delivery and execution, on the premise that AI-generated code runs itself on targets.
b) It lowers the skill floor and speeds iteration, but the code still needs delivery, execution and persistence.
c) It guarantees that the resulting malware will evade all behavioral detection once it runs.
d) It has no effect at all, because models cannot actually produce code.
10. A stakeholder proposes that the organization's sole AI-misuse defense should be contractually requiring staff to only use models that "cannot be jailbroken."
What is the strongest objection?
a) Such perfectly unbreakable models already exist and are entirely safe to use, so the only real issue left here is their cost.
b) The objection is that jailbreaking is impossible, which would make the contract clause simply redundant.
c) No refusal training is guaranteed unbreakable, so a policy premised on one is unsound; layered controls are needed.
d) There is no real objection, since a contract clause fully removes the technical risk.
Answers:
|
Question: 01 Answer: c |
Question: 02 Answer: a, d |
Question: 03 Answer: b |
Question: 04 Answer: d |
Question: 05 Answer: a |
|
Question: 06 Answer: d |
Question: 07 Answer: c |
Question: 08 Answer: a |
Question: 09 Answer: b |
Question: 10 Answer: c |
Note: For any error in GIAC Offensive AI Analyst (GOAA) certification exam sample questions, please update us by writing an email on feedback@edusum.com.
