secml_malware.attack.blackbox.ga package

Submodules

secml_malware.attack.blackbox.ga.c_base_genetic_engine module

class secml_malware.attack.blackbox.ga.c_base_genetic_engine.CGeneticAlgorithm(problem: secml_malware.attack.blackbox.c_blackbox_problem.CBlackBoxProblem, is_debug: bool = False)

Bases: secml.adv.attacks.evasion.c_attack_evasion.CAttackEvasion

The genetic optimizer.

f_eval()

Returns the number of function evaluations made during the attack.

grad_eval()

Returns the number of gradient evaluations made during the attack.

objective_function(x)

Objective function.

Parameters

x (CArray or CDataset) –

Returns

f_obj

Return type

float or CArray of floats

objective_function_gradient(x)

Gradient of the objective function.

run(x, y, ds_init=None) → Tuple[secml.array.c_array.CArray, secml.array.c_array.CArray, secml.data.c_dataset.CDataset, Any]

Runs the genetic algorithms.

Parameters

  • x (CArray) – input sample to perturb

  • y (CArray) – original class

  • ds_init (CDataset, optional, default None) – the initialization point. Default is None

Returns

  • CArray – y_pred : the predicted label after the attack

  • CArray – scores : the scores after the attack

  • CDataset – adv_ds : the CDataset containing the adversarial points

  • CArray – f_obj : the mean value for the objective function

classmethod write_adv_to_file(x_adv: secml.array.c_array.CArray, path: str)

Write the adversarial malware as a file on disk

Parameters

  • x_adv (CArray) – The adversarial malware to dump

  • path (str) – The path where to save the executable

secml_malware.attack.blackbox.ga.c_base_genetic_engine.random_mutation(individual, indpb)

Apply the mutation operator, that perturb randomly each entry of the individual, with a given probability. The mutation is applied in-place.

Parameters

  • individual – the individual to mutate

  • indpb (float) – the probability of altering a single entry

Returns

the mutated individual, the mutatio is in-place

Return type

tuple

Module contents