AAIA: Advanced in AI Audit

By Isaca · Question Mode

QuestionQ9

AI Governance and Risk

An organization deploys an AI-driven image-recognition system that is susceptible to evasion attacks. Which approach BEST helps ensure that the system mitigates these evasion attempts?

Explanation

Adversarial training incorporates adversarial examples into model training so the model learns to classify inputs correctly despite malicious, small perturbations designed to evade detection. This directly improves resilience to evasion attacks.

Community Discussion

No comments yet. Be the first to start the discussion!