InverseNeuralNetwork/test.py at master · irijije/InverseNeuralNetwork · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
import os
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt

from preprocess import load_dataset
from InverseNN import InverseNN
from InverseNN_add import InverseNN_add


os.environ["CUDA_VISIBLE_DEVICES"] = '1'
NN_MODEL_NAME = "models/nn.h5"
AE_ENCODER_NAME = "models/ae_encoder.h5"
INN_MODEL_NAME = "models/inv_small_test.h5"
TEST_DATA = "dataset/CICIDS2018_mal_test.csv"

BLACKBOX = True


def show_test():
    x, _ = load_dataset("dataset/test.csv")
    x = np.array(x)
    nn_model = tf.keras.models.load_model(NN_MODEL_NAME)
    encoder = tf.keras.models.load_model(AE_ENCODER_NAME)
    inn_model = tf.keras.models.load_model(INN_MODEL_NAME)

    if BLACKBOX:
        y = nn_model(x, training=False)
        max_y = tf.argmax(y, axis=1)
        onehot_y = tf.reshape(tf.one_hot(max_y, depth=15), shape=[-1, 15])
        x_f = encoder(x, training=False)
        x_ = inn_model(tf.concat([onehot_y, x_f], axis=1), training=False)
        y_ = nn_model(x_, training=False)
    else:
        y = nn_model(x, training=False)
        x_ = inn_model(y, training=False)
        y_ = nn_model(x_, training=False)

    print("\nx: {}".format(np.array2string(np.array(x), prefix="x: ",
            formatter={'float_kind':lambda x: "%7.4f" % x})))
    print("\ny: {}".format(np.array2string(np.array(y), prefix="y: ",
            formatter={'float_kind':lambda x: "%7.4f" % x})))
    print("\nx_: {}".format(np.array2string(np.array(x_), prefix="x_: ",
            formatter={'float_kind':lambda x: "%7.4f" % x})))
    print("\ny_: {}".format(np.array2string(np.array(y_), prefix="y_: ",
            formatter={'float_kind':lambda x: "%7.4f" % x})))

    y = np.argmax(y)
    y_ = np.argmax(y_)

    if y != y_:
        print("\nattack success y: {} y_: {} \n".format(y, y_))
    else:
        print("\nattack failed y: {} y_: {} \n".format(y, y_))

def test(filename=INN_MODEL_NAME):
    x, _ = load_dataset(TEST_DATA)
    x = np.array(x)
    nn_model = tf.keras.models.load_model(NN_MODEL_NAME)
    encoder = tf.keras.models.load_model(AE_ENCODER_NAME)
    inn_model = tf.keras.models.load_model(filename)

    if BLACKBOX:
        y = nn_model(x, training=False)
        # max_y = tf.argmax(y, axis=1)
        # onehot_y = tf.reshape(tf.one_hot(max_y, depth=15), shape=[-1, 15])
        # x_f = encoder(x, training=False)
        #x_ = inn_model(tf.concat([onehot_y, x_f], axis=1), training=False)
        x_ = inn_model(y, training=False)
        y_ = nn_model(x_, training=False)
    else:
        y = nn_model(x, training=False)
        x_ = inn_model(y, training=False)
        y_ = nn_model(x_, training=False)

    y = tf.math.argmax(y, axis=1)
    y_ = tf.math.argmax(y_, axis=1)

    #print(y)
    #print(y_)

    generate_loss = tf.reduce_mean(tf.square(tf.abs(x - x_)))

    acc = tf.keras.metrics.Accuracy()
    acc.update_state(y_, tf.zeros_like(y))
    print("attack success probability: {}".format(acc.result().numpy()))

    return generate_loss, acc.result().numpy()


if __name__ == "__main__":
    losses = []
    success_rate = []
    for i in range(21):
        C = 2**-(20-i)
        if BLACKBOX:
            model_name = "models/inv_add_my2"+str(C)+".h5"
            inn = InverseNN_add()
        else:
            model_name = "models/inv_cw_"+str(C)+".h5"
            inn = InverseNN()
        print("model {}".format(model_name))
        inn.train(C, 3, model_name)
        loss, rate = test(model_name)
        losses.append(loss)
        success_rate.append(rate)

    superscript_map = {
    "0": "⁰", "1": "¹", "2": "²", "3": "³", "4": "⁴", "5": "⁵", "6": "⁶",
    "7": "⁷", "8": "⁸", "9": "⁹", "-": "⁻"}

    trans = str.maketrans(''.join(superscript_map.keys()), ''.join(superscript_map.values()))

    plt.title("Evasion Attack with Autoencoder")
    loss_ax = plt.gca()
    xs = ["2"+("-"+str(20-x)).translate(trans) for x in range(21)]
    rate_ax = loss_ax.twinx()
    rate_ax.set_ylim(0.0, 1.05)
    loss_ax.set_ylim(0.0, 0.0105)
    rate_ax.set_yticks(np.arange(0, 1.1, 0.1))
    loss_ax.set_yticks(np.arange(0, 0.011, 0.001))
    l1, = rate_ax.plot(xs, success_rate, 'C0', marker='.')
    l2, = loss_ax.plot(xs, losses, 'C1', marker='.')
    loss_ax.set_xlabel('C values')
    rate_ax.set_ylabel('attack success rate')
    loss_ax.set_ylabel('generate loss')
    plt.legend([l1, l2], ["rate", "loss"], loc='upper left')
    plt.grid(b=True, which='major', linestyle='--')
    plt.savefig('c_values.png')
    plt.show()

    #test()
    #show_test()