Return the absolute file of the 'filename' helper file :param parameters: :param filename: file name :return:

set the specified parameter to the specified value and write back to the *.ini file :param parameters_filepath: filename (absolute path)

writes a df to tex file :param df: dataframe to be converted into tex table :param filepath: tex filepath :return:

ORs two float tensors by converting them to byte and back Note that byte() takes the first 8 bit after the decimal point of the float e.g

:return: the ratio of the visited samples to the maximum expected ones

Helper function to randomly initialize the the inner maximizer algos randomize such that the functionality is preserved. Functionality is

Stack arbitrary number of tensors along the first dimension :param args: list of tensors :return: tensor stacking all the input tensors

load parameters from an *.ini file :param parameters_filepath: filename (absolute path) :return: nested dictionary of parameters

Function to be used for both testing and validation :param epoch: current epoch :param is_validate: is the testing done on the validation

test the model in terms of loss and accuracy on category, this function also allows to perform perturbation with respect to loss to evade

Clip a tensor to be within lb and ub :param x: :param lb: lower bound (scalar) :param ub: upper bound (scalar) :return: clipped v

Load the training/test datasets :param parameters: :return: dictionaries of train and test dataloaders

rounds x by thresholding it according to alpha which can be a scalar or vector :param x: :param alpha: threshold parameter :return: a

update the covering number measure with the new batch :param batch_idx: current batch index :param batch: batch features in t

XORs two float tensors by converting them to byte and back Note that byte() takes the first 8 bit after the decimal point of the float e.

Helper function to get the list of imported function calls for a binary :param filepath: binary's absolute filepath :retu

assign a rank to the file can be used for sorting :param filename: :return:

A wrapper function for the above algorithim :param iterations: :param x: :param y: :param model: :param loss_fct: :param

Multi-step bit coordinate ascent :param use_sample: :param is_report_loss_diff: :param y: :param x: (tensor) feature vector :

Multi-step bit gradient ascent :param x: (tensor) feature vector :param y: :param model: nn model :param loss_fct: loss function

Constructs a neural net binary classifer :param input_size: :param hidden_1_size: :param hidden_2_size: :param hidden_3_size:

Creates mapping of all the lib imports within benign and malicious samples into their corresponding indices in the feature vector. The mapp

Create TeX tables from the results populated under `result_files` which is generated from running `framework.py` The tex file is stored i

FGSM^k with deterministic rounding :param y: :param x: (tensor) feature vector :param model: nn model :param loss_fct: loss funct

Multi-step bit coordinate ascent using gradient of output, advancing in direction of maximal change :param use_sample: :param is_report_l

FGSM^k with randomized rounding :param x: (tensor) feature vector :param y: :param model: nn model :param loss_fct: loss function

update the numerator counter for the sample_idxth point by testing if point has already been visited :param sample_idx: index of the

file_abs_locations: PE file names including path to is_malicious: either benign or malicious files in a single dataset

initialize the weights of a network :param net: :return:

Computes the denominator of the sample_idxth sample of the training data. This method needs to be called once as the denominator is c