Which defense is MOST effective against cyberattacks that alter input data to avoid detection?

The correct answer is A. Enhancing model robustness through adversarial training. Attacks that alter input data to evade detection are adversarial examples - inputs crafted to cause misclassification. Adversarial training (A) directly counters this by exposing the model to adversarial examples during training, teaching it to correctly classify perturbed…

AI Security Design and Implementation

Question

Options

AEnhancing model robustness through adversarial training
BRestricting access to internal model parameters
CConducting periodic monitoring of decisions
DApplying differential privacy to training data

How the community answered

(37 responses)

A
70% (26)
B
5% (2)
C
8% (3)
D
16% (6)

Explanation

Attacks that alter input data to evade detection are adversarial examples - inputs crafted to cause misclassification. Adversarial training (A) directly counters this by exposing the model to adversarial examples during training, teaching it to correctly classify perturbed inputs. This hardens the model's decision boundaries against such manipulation. Restricting access to model parameters (B) addresses model extraction threats, not input manipulation. Periodic monitoring (C) detects drift after the fact but does not prevent adversarial bypass. Differential privacy (D) protects training data confidentiality but does not make the model resistant to adversarial inputs at inference time.

Topics

#Adversarial attacks#Model robustness#Adversarial training#Input manipulation

Community Discussion

No community discussion yet for this question.

Full AAISM Practice