An example of an adversarial attack is?

• A. Misinterpreting A Stop Sign Due To Crafted Input
• B. Data Poisoning Of Training Data
• C. Privacy Breach
• D. Insufficient Explainability

Answer: (A)

Adversarial attacks involve feeding a model inputs that have been carefully altered to cause it to make a wrong decision, even though a human would see the input as legitimate. A stop sign that is misinterpreted due to crafted input is a quintessential example: the image is subtly perturbed so the model’s vision system misclassifies it, while the disruption remains almost invisible to a human observer. This demonstrates a vulnerability at inference time where the model can be tricked by deliberate perturbations. Data poisoning of training data is related but occurs during the learning phase, aiming to influence future predictions by tampering with the training set rather than manipulating a single input at run time. Privacy breach and insufficient explainability involve different security or governance concerns—data access control and transparency—rather than directly deceiving the model with a crafted input.