network.py

import itertools
import logging
import os
import time
from pathlib import Path

import attr
import matplotlib.pyplot as plt
import numpy as np
import torch
from torch import nn
from torch import optim

from src.config import NeuralNetworkConfig, RootConfig
from src.data.loader import np_matrix_from_scored_csv

logger = logging.getLogger(__name__)
DEVICE = torch.device(os.environ.get('device', 'cuda'))

@attr.s
class FormattedData:
    unique_x: np.array = attr.ib()
    probs_x: np.array = attr.ib()
    y: np.array = attr.ib()


@attr.s
class BenignData:
    unique_x: np.array = attr.ib()
    counts: np.array = attr.ib()
    y: np.array = attr.ib()

class OrderCounter:
    order = 0

    @staticmethod
    def next():
        OrderCounter.order += 1
        return OrderCounter.order


class SoftClip(nn.Module):
    """ SoftClipping activation function

    https://arxiv.org/pdf/1810.11509.pdf
    """
    def __init__(self, p=50.0):
        super().__init__()
        self.p = p

    def forward(self, x):
        first_pow = torch.pow(np.e, torch.mul(x, self.p))
        second_pow = torch.pow(np.e, torch.mul(torch.add(x, -1.0), self.p))

        first_div = torch.add(first_pow, 1.0)
        second_div = torch.add(second_pow, 1.0)

        division = torch.div(first_div, second_div)

        second_part_log = torch.log(division)
        first_part = 1.0 / self.p

        if len(second_part_log[torch.isnan(second_part_log)]):
            print("prdel")  # todo remove me

        return torch.mul(second_part_log, first_part)


class NeuralNetwork:
    def __init__(self, input_features=2,
                 nn_conf: NeuralNetworkConfig = NeuralNetworkConfig()):
        self.model = nn.Sequential(
            nn.Linear(input_features, 20),
            nn.ReLU(),
            nn.Linear(20, 15),
            nn.ReLU(),
            nn.Linear(15, 20),
            nn.ReLU(),
            nn.Linear(20, 1),
            nn.Tanh(),
            SoftClip(50)
            # nn.Sigmoid()
        ).to(DEVICE)
        self._set_weights()
        self.conf = nn_conf
        self.id = OrderCounter.next()

        # Variables used for loss function
        self.attacker_actions: FormattedData = None
        self.benign_data: BenignData = None

        # Variables from last training epoch measuring quality
        self.final_loss = None
        self.final_fp_cost = None

        # -------------- <TMP> --------------------
        # self.max_constant = constant
        # self.cur_value = .0
        # self.step = .05
        # self.edge_epoch = (self.conf.epochs*0.2)
        # self.incr_each = self.edge_epoch / (self.max_constant/.05)

    # def get_cur_coefficient(self, epoch) -> float:
    #     return self.max_constant
        # if epoch > self.edge_epoch:
        #     return self.max_constant
        # return .01
        # if self.cur_value < self.max_constant and epoch % self.incr_each == 0:
        #     self.cur_value += self.step
        # return self.cur_value
    # -------------- </TMP> --------------------

    def __str__(self):
        return f'Neural network id:{self.id}, final loss: {self.final_loss}'

    def set_data(self, benign_data: BenignData, attack: FormattedData):
        self.attacker_actions = attack
        self.benign_data = benign_data

    def _prepare_data(self):
        defender = self.benign_data
        attacker = self.attacker_actions
        x = np.concatenate((defender.unique_x, attacker.unique_x), axis=0)
        y = np.concatenate((defender.y, attacker.y), axis=0)
        probs = np.concatenate((defender.counts/np.sum(defender.counts),
                                attacker.probs_x), axis=0)

        self.train_y = torch.cat((torch.zeros(self.conf.batch_size).float(),
                                  torch.tensor(attacker.y).float())).to(DEVICE)

        self.all_x = torch.tensor(x).float().to(DEVICE)
        self.all_y = torch.tensor(y).float().to(DEVICE)
        self.all_probs = torch.tensor(probs).float().to(DEVICE)

    def _set_weights(self):
        def init_weights(m):
            if type(m) == nn.Linear:
                torch.nn.init.xavier_uniform_(m.weight)
                m.bias.data.fill_(.0)
        self.model.apply(init_weights)

    def train(self):
        self._prepare_data()
        self._train()

    def get_train_batch(self):
        batch_idxs = np.random.choice(len(self.benign_data.unique_x),
                                      self.conf.batch_size)
        current_batch_samples = np.sum(self.benign_data.counts[batch_idxs])

        batch_x_np = np.concatenate((self.benign_data.unique_x[batch_idxs],
                                     self.attacker_actions.unique_x), axis=0)
        batch_probs_np = np.concatenate((self.benign_data.counts[batch_idxs]/current_batch_samples,
                                 self.attacker_actions.probs_x), axis=0)

        batch_x = torch.tensor(batch_x_np).float().to(DEVICE)
        batch_probs = torch.tensor(batch_probs_np).float().to(DEVICE)
        return batch_x, batch_probs

    def _train(self):
        learning_rate = self.conf.learning_rate
        optimizer = torch.optim.Adam(self.model.parameters(), lr=learning_rate)

        for e in range(self.conf.epochs):
            batch_x, batch_probs = self.get_train_batch()

            # Forward pass: compute predicted y by passing x to the model
            train_ltncies = self.latency_predict(batch_x, with_grad=True)

            # Compute loss
            loss, _ = self.conf.loss_function(batch_x, train_ltncies,
                                              self.train_y, batch_probs)

            # Log loss function value each 5 epochs
            if e % 50 == 0:
                logging.debug(f'Epoch: {e}/{self.conf.epochs},\t'
                              f'TrainLoss: {loss},\t')

            # Before the backward pass, use the optimizer object to zero all of
            # the gradients for the variables it will update
            optimizer.zero_grad()

            # Backward pass: compute gradient of the loss with respect to model
            # parameters
            loss.backward()

            # Calling the step function on an Optimizer makes an update to its
            # parameters
            optimizer.step()

        with torch.no_grad():
            train_ltncies = self.latency_predict(self.all_x, with_grad=True)
            loss, fp_part = self.conf.loss_function(self.all_x, train_ltncies,
                                                 self.all_y, self.all_probs)
            logger.debug(f'Final loss of this nn: {loss}\tfp_part is: {fp_part}')
            # measuring quality of final network
            self.final_loss = loss.item()
            self.final_fp_cost = fp_part.item()

    def plot_tmp(self):
        # todo remove me
        if self.id == 1:
            return
        try:
            actions = self.actions
        except AttributeError:
            self.plotted = []
            self.plotting = plt.subplots()
            one_axis = np.linspace(0, 1, 101)  # [0.00, 0.01, 0.02, ..., 0.99, 1.00]
            generator = itertools.product(*itertools.repeat(one_axis, 2))
            actions = torch.tensor(np.array(list(generator))).float().to(DEVICE)
            self.actions = actions
        finally:
            # Remove all lines from previous iteration plotting
            for item in self.plotted:
                item.remove()
            self.plotted = []
            res = self.latency_predict(actions).numpy().reshape((101, 101), order='F')
            self.plotted.append(self.plotting[1].imshow(res, cmap='Reds', vmin=0, vmax=1, origin='lower'))
            self.plotting[0].canvas.draw()
            plt.pause(0.000001)
            time.sleep(1)

    def _raw_predict(self, tensor: torch.Tensor):
        pred = self.model(tensor)
        return pred.flatten().float()

    def latency_predict(self, x: torch.Tensor, with_grad=False):
        if with_grad:
                raw_prediction = self._raw_predict(x)
        else:
            with torch.no_grad():
                raw_prediction = self._raw_predict(x)

        return raw_prediction

    def predict_single_latency(self, input, return_tensor=False):
        in_type = type(input)
        if in_type == list or in_type == tuple or \
                in_type == np.array or in_type == np.ndarray:
            input = torch.tensor(input).float().to(DEVICE)

        if return_tensor:
            return self.latency_predict(input)[0]
        else:
            return self.latency_predict(input)[0].item()


def setup_loger(debug: bool):
    log_format = ('%(asctime)-15s\t%(name)s:%(levelname)s\t'
                  '%(module)s:%(funcName)s:%(lineno)s\t%(message)s')
    level = logging.DEBUG if debug else logging.INFO
    logging.basicConfig(level=level, format=log_format)


if __name__ == '__main__':
    setup_loger(True)
    benign_x, _ = np_matrix_from_scored_csv(
        Path('all_benign_scored.csv'), 0, 500)
    malicious_x, _ = np_matrix_from_scored_csv(
        Path('scored_malicious.csv'), 1, 400)

    benign_unique_x, counts = np.unique(benign_x, axis=0, return_counts=True)
    probs_benign = np.array([count / len(benign_x) for count in counts])
    benign_y = np.zeros(len(benign_unique_x))
    benign_data = FormattedData(benign_unique_x, probs_benign, benign_y)

    malicious_unique_x, counts = np.unique(malicious_x, axis=0, return_counts=True)
    probs_malicious = np.array([count / len(malicious_unique_x) for count in counts])
    malicious_y = np.ones(len(malicious_unique_x))
    malicious_data = FormattedData(malicious_unique_x, probs_malicious, malicious_y)

    conf = RootConfig()
    nn = NeuralNetwork(2, conf.model_conf.defender_conf.nn_conf)
    nn.set_data(benign_data, malicious_data)
    nn.train()