CWAttack<T, TInput, TOutput>
Models & Types · AiDotNet.AdversarialRobustness.Attacks
Implements the Carlini and Wagner (C and W) attack.
Inheritance: AdversarialAttackBase<T, TInput, TOutput> → CWAttack<T, TInput, TOutput>
For Beginners
C and W is one of the most sophisticated attacks. Instead of following gradients, it treats creating adversarial examples as a carefully crafted optimization problem. It’s slower than FGSM or PGD but often finds adversarial examples that are more subtle and harder to defend against.
How It Works
C and W is an optimization-based attack that formulates adversarial example generation as an optimization problem, typically producing stronger attacks than gradient-based methods.
Original paper: “Towards Evaluating the Robustness of Neural Networks” by Carlini and Wagner (2017)
Constructors
| Constructor | Summary |
|---|---|
CWAttack(AdversarialAttackOptions<T> options) |
Initializes a new instance of the C and W attack. |
Methods
| Method | Summary |
|---|---|
TInput CalculatePerturbation(TInput original, TInput adversarial) |
|
TInput GenerateAdversarialExample(TInput input, TOutput trueLabel, IFullModel<T, TInput, TOutput> targetModel) |
Generates an adversarial example using the C and W L2 attack. |