secml.ml.classifiers.secure

CClassifierSecSVM

class secml.ml.classifiers.secure.c_classifier_sec_svm.CClassifierSecSVM(ub=inf, idx_ub=None, lb=-inf, idx_lb=None, eta=0.5, max_it=10000.0, eps=0.0001, *args, **kwargs)

Bases: secml.ml.classifiers.sklearn.c_classifier_svm.CClassifierSVM

Secure Support Vector Machine (Sec-SVM) classifier.

This implements the secure classifier from:

Demontis et al. “Yes, machine learning can be more secure! a case study on android malware detection.” IEEE TDSC 2017. https://arxiv.org/abs/1704.08996

Parameters

ubscalar or None, optional

Upper bound of the weights. If None (default), no bound is applied.

idx_ubCArray or None, optional

If CArray, the upper bound is only applied to the weights indicized by idx_ub. If None (default), the bound is applied to all weights.

lbscalar or None, optional

Lower bound of the weights. If None (default), no bound is applied.

idx_lbCArray or None, optional

If CArray, the lower bound is only applied to the weights indicized by idx_ub. If None (default), the bound is applied to all weights.

etascalar, optional

Step of the gradient descent. Default 0.5.

max_itint, optional

Maximum number of iterations of the gradient descent. Default 1e4.

epsscalar, optional

Tolerance of the stop criterion of the gradient descent. Default 1e-4.

*args, **kwargs

Other paramters from CClassifierSVM.

Attributes

class_type‘sec-svm’

Defines class type.

Methods

C_hinge_loss(self, x, y)	Compute the loss term for each point in dataset multiplied by C.
backward(self[, w])	Returns the preprocessor gradient wrt data.
copy(self)	Returns a shallow copy of current class.
create([class_item])	This method creates an instance of a class with given type.
create_chain(class_items, kwargs_list)	Creates a chain of preprocessors.
decision_function(self, x[, y])	Computes the decision function for each pattern in x.
deepcopy(self)	Returns a deep copy of current class.
estimate_parameters(self, dataset, …[, …])	Estimate parameter that give better result respect a chose metric.
fit(self, x, y)	Trains the classifier.
fit_forward(self, x[, y, caching])	Fit estimator using data and then execute forward on the data.
forward(self, x[, caching])	Forward pass on input x.
get_class_from_type(class_type)	Return the class associated with input type.
get_params(self)	Returns the dictionary of class hyperparameters.
get_state(self, **kwargs)	Returns the object state dictionary.
get_subclasses()	Get all the subclasses of the calling class.
grad_f_params(self, x[, y])	Derivative of the decision function w.r.t.
grad_f_x(self, x, y)	Computes the gradient of the classifier’s decision function wrt x.
grad_loss_params(self, x, y[, loss])	Derivative of the loss w.r.t.
grad_tr_params(self, x, y)	Derivative of the classifier training objective w.r.t.
gradient(self, x[, w])	Compute gradient at x by doing a backward pass.
gradient_w_b(self, x, y)	Compute the gradient dloss/dw, where loss is sum_i max(0, 1-y_i*f(x_i))
hessian_tr_params(self[, x, y])	Hessian of the training objective w.r.t.
hinge_loss(self, x, y)	Compute the loss term for each point in dataset.
is_fitted(self)	Return True if the classifier is trained (fitted).
list_class_types()	This method lists all types of available subclasses of calling one.
load(path)	Loads object from file.
load_state(self, path)	Sets the object state from file.
objective(self, x, y)	Objective function.
predict(self, x[, return_decision_function])	Perform classification of each pattern in x.
save(self, path)	Save class object to file.
save_state(self, path, **kwargs)	Store the object state to file.
set(self, param_name, param_value[, copy])	Set a parameter of the class.
set_params(self, params_dict[, copy])	Set all parameters passed as a dictionary {key: value}.
set_state(self, state_dict[, copy])	Sets the object state using input dictionary.
timed([msg])	Timer decorator.

C_hinge_loss(self, x, y)

Compute the loss term for each point in dataset multiplied by C.

If class_weight == ‘balanced’, it multiplies C to the inverse prob of theclasses.

property b

Return the SVM bias (b term in the decision function)

property eps

Precision of the stop condition for training.

property eta

Eta parameter for the training gradient.

gradient_w_b(self, x, y)

Compute the gradient dloss/dw, where loss is sum_i max(0, 1-y_i*f(x_i))

hinge_loss(self, x, y)

Compute the loss term for each point in dataset.

property lb

Return value of weight lower bound

property max_it

Maximum number of iteration for the training.

objective(self, x, y)

Objective function.

property ub

Return value of weight upper bound

property w

Return the vector of feature weights (only if ckernel is None)