Which of the following BEST describes an adversarial attack on an AI model?

The correct answer is B. Providing inputs that mislead the model into incorrect predictions. Adversarial attacks involve crafting specially manipulated inputs-often imperceptible to humans-that exploit the mathematical properties of a model to cause it to produce incorrect or attacker-desired predictions. Classic examples include perturbing image pixels to fool an…

AI Security Risk Management

Question

Options

AAttacking underlying hardware
BProviding inputs that mislead the model into incorrect predictions
CReverse-engineering the model using social engineering
DConducting denial-of-service attacks on AI APIs

How the community answered

(46 responses)

A
4% (2)
B
89% (41)
C
4% (2)
D
2% (1)

Explanation

Adversarial attacks involve crafting specially manipulated inputs-often imperceptible to humans-that exploit the mathematical properties of a model to cause it to produce incorrect or attacker-desired predictions. Classic examples include perturbing image pixels to fool an image classifier or crafting text prompts to bypass content filters. Option A (hardware attacks) targets physical infrastructure, not the model itself. Option C conflates social engineering (a human manipulation tactic) with model reverse-engineering-these are separate concepts. Option D (DoS on APIs) is a availability attack against infrastructure, not an adversarial manipulation of model behavior.

Topics

#Adversarial AI#AI model attacks#Input manipulation#Model robustness

Community Discussion

No community discussion yet for this question.

Full AAISM Practice