advsecurenet.attacks.decision_based package

---

advsecurenet.attacks.decision_based.boundary module

class advsecurenet.attacks.decision_based.boundary.DecisionBoundary(config: DecisionBoundaryAttackConfig)

Bases: AdversarialAttack

Decision Boundary attack

Paper: https://arxiv.org/abs/1712.04248

attack(model: BaseModel, x: Tensor, y: Tensor) → Tensor

Generates adversarial examples using the Decision Boundary attack.

Parameters:

• model (BaseModel) – The model to attack.

• x (torch.Tensor) – The original images to attack. Expected shape is (batch_size, channels, height, width).

• y (torch.Tensor) – If the attack is untargeted, the true labels for the images. If the attack is targeted, the target labels for the images. Expected shape is (batch_size,).

Returns:

The adversarial images.

Return type:

torch.Tensor

---