secmlt.trackers package

Submodules

secmlt.trackers.image_trackers module

Image-specific trackers.

class secmlt.trackers.image_trackers.GradientsTracker

Bases: Tracker

Tracker for gradient images.

__init__() → None

Create gradients tracker.

track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.Tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Track the gradients at the current iteration as images.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

class secmlt.trackers.image_trackers.SampleTracker

Bases: Tracker

Tracker for adversarial images.

__init__() → None

Create adversarial image tracker.

track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.Tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Track the adversarial examples at the current iteration as images.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

secmlt.trackers.tensorboard_tracker module

Tensorboard tracking utilities.

class secmlt.trackers.tensorboard_tracker.TensorboardTracker(logdir: str, trackers: list[Tracker] | None = None)

Bases: Tracker

Tracker for Tensorboard. Uses other trackers as subscribers.

__init__(logdir: str, trackers: list[Tracker] | None = None) → None

Create tensorboard tracker.

Parameters:

• logdir (str) – Folder to store tensorboard logs.

• trackers (list[Tracker] | None, optional) – List of trackers subsctibed to the updates, by default None.

get_last_tracked() → NotImplementedError

Not implemented for this tracker.

track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Update all subscribed trackers.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

secmlt.trackers.trackers module

Trackers for attack metrics.

class secmlt.trackers.trackers.GradientNormTracker(p: LpPerturbationModels = 'l2')

Bases: Tracker

Tracker for gradients.

__init__(p: LpPerturbationModels = 'l2') → None

Create gradient norm tracker.

Parameters:

p (LpPerturbationModels, optional) – Perturbation model to compute the norm, by default LpPerturbationModels.L2.

track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Track the sample-wise gradient of the loss w.r.t delta.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

class secmlt.trackers.trackers.LossTracker

Bases: Tracker

Tracker for attack loss.

__init__() → None

Create loss tracker.

track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Track the sample-wise loss of the attack at the current iteration.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

class secmlt.trackers.trackers.PerturbationNormTracker(p: LpPerturbationModels = 'l2')

Bases: Tracker

Tracker for perturbation norm.

__init__(p: LpPerturbationModels = 'l2') → None

Create perturbation norm tracker.

Parameters:

p (LpPerturbationModels, optional) – Perturbation model to compute the norm, by default LpPerturbationModels.L2.

track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Track the perturbation norm at the current iteration.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

class secmlt.trackers.trackers.PredictionTracker

Bases: Tracker

Tracker for model predictions.

__init__() → None

Create prediction tracker.

track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Track the sample-wise model predictions at the current iteration.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

class secmlt.trackers.trackers.ScoresTracker(y: int | torch.Tensor | None = None)

Bases: Tracker

Tracker for model scores.

__init__(y: int | torch.Tensor | None = None) → None

Create scores tracker.

track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Track the sample-wise model scores at the current iteration.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

class secmlt.trackers.trackers.Tracker(name: str, tracker_type: str = 'scalar')

Bases: ABC

Class implementing the trackers for the attacks.

__init__(name: str, tracker_type: str = 'scalar') → None

Create tracker.

Parameters:

• name (str) – Tracker name.

• tracker_type (str, optional) – Type of tracker (mostly used for tensorboard functionalities), by default SCALAR. Available: SCALAR, IMAGE, MULTI_SCALAR.

get() → torch.Tensor

Get the current tracking history.

Returns:

History of tracked parameters.

Return type:

torch.Tensor

get_last_tracked() → None | torch.Tensor

Get last element tracked.

Returns:

Returns the last tracked element if anything was tracked.

Return type:

None | torch.Tensor

abstract track(iteration: int, loss: torch.Tensor, scores: torch.Tensor, x_adv: torch.tensor, delta: torch.Tensor, grad: torch.Tensor) → None

Track the history of given attack observable parameters.

Parameters:

• iteration (int) – The attack iteration number.

• loss (torch.Tensor) – The value of the (per-sample) loss of the attack.

• scores (torch.Tensor) – The output scores from the model.

• x_adv (torch.tensor) – The adversarial examples at the current iteration.

• delta (torch.Tensor) – The adversarial perturbations at the current iteration.

• grad (torch.Tensor) – The gradient of delta at the given iteration.

Module contents

Module implementing trackers for adversarial attacks.