IoT-Malware-Detection-Solution-Using-CNN-Classification/write_hand_Model.py at main · nmthuann/IoT-Malware-Detection-Solution-Using-CNN-Classification · 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

import cv2
import numpy as np
import matplotlib.pyplot as plt

# lay so dep
# np.random.seed(5)

img = cv2.imread('test_image_cnn.jpg')
img = cv2.resize(img, (200,200))
img_gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)/255 # sinh gia tri lớn phải chia để chuẩn hóa dữ liệu -> tràn bộ nhớ
# print(img_gray)

#thiet ke convolution layer
class Conv2d:
    def __init__(self, input, numOfKernel = 8, kernel_size = 3, padding=0, stride=1):
        self.input = np.pad(input, ((padding, padding),(padding, padding)), 'constant')
        self.stride = stride
        # self.height, self.width = input.shape
        self.kernel = np.random.randn(numOfKernel, kernel_size, kernel_size)
        # print(kernel)

        self.results = np.zeros((int((self.input.shape[0] - self.kernel.shape[1])/self.stride) + 1,
                                 int((self.input.shape[1] - self.kernel.shape[2])/self.stride) + 1,
                                self.kernel.shape[0]))

    # roi: region of interest
    def getROI(self):
        for row in range(int((self.input.shape[0] - self.kernel.shape[1])/self.stride) + 1):
            for col in range(int((self.input.shape[1] - self.kernel.shape[2])/self.stride) + 1):
                roi = self.input[row*self.stride: row*self.stride + self.kernel.shape[1],
                      col*self.stride: col*self.stride + self.kernel.shape[2]]
                yield row, col, roi

    # toán tích chập
    def operate(self):
        for layer in range(self.kernel.shape[0]):
            for row, col, roi in self.getROI():
              self.results[row, col, layer] = np.sum(roi * self.kernel[layer])

        return self.results

class Relu:
    def __init__(self, input):
        self.input = input
        self.results = np.zeros((self.input.shape[0],
                                 self.input.shape[1],
                                 self.input.shape[2])) # định trước vùng chứa

    def operate(self):
        for layer in range(self.input.shape[2]):
            for row in range(self.input.shape[0]):
                for col in range(self.input.shape[1]):
                    self.results[row, col, layer] = 0 if self.input[row, col, layer] < 0 else self.input[row, col, layer]

        return self.results

class LeakyRelu:
    def __init__(self, input):
        self.input = input
        self.results = np.zeros((self.input.shape[0], self.input.shape[1])) # định trước vùng chứa

    def operate(self):
        for row in range(self.input.shape[0]):
            for col in range(self.input.shape[1]):
                self.results[row, col] = 0.1*self.input[row, col] if self.input[row, col] < 0 else self.input[row, col]

        return self.results


class MaxPooling:
    def __init__(self, input, poolingSize):
        self.input = input
        self.poolingSize = poolingSize

        self.result = np.zeros((int(self.input.shape[0]/self.poolingSize),
                                int(self.input.shape[1]/self.poolingSize),
                                 self.input.shape[2]))  # định trước vùng chứa

    def operate(self):
        for layer in range(self.input.shape[2]):
            for row in range(int(self.input.shape[0]/self.poolingSize)):
                for col in range(int(self.input.shape[1]/self.poolingSize)):
                    self.result[row, col, layer] = np.max(self.input[row*self.poolingSize: row*self.poolingSize+self.poolingSize,
                        col * self.poolingSize: col * self.poolingSize + self.poolingSize, layer])

        return self.result

img_gray_conv2d = Conv2d(img_gray, 16, 3, padding=0, stride=1).operate()
img_gray_conv2d_relu = Relu(img_gray_conv2d).operate()
img_gray_conv2d_relu_maxpooling = MaxPooling(img_gray_conv2d_relu,3).operate()


fig = plt.figure(figsize=(10,10))
for i in range(16):
    plt.subplot(4,4, i + 1)
    plt.imshow(img_gray_conv2d_relu[:,:,i], cmap='gray')
    plt.axis('off')
#plt.savefig('img_gray_conv2d.jpg')
plt.show()


# # entropy - cross entropy - KL Divergence
# # entropy: mức độ hỗn loạn không đồng nhất của 1 mẫu nào đó.
# # entropy = 0: là đồng nhất => nguyên chất
# # tính entropy: entropy = -sum(P(i)*loge(P(i))) cơ số e => mức độ ko đồng nhất
# # sự đồng nhất: p [1,0,0], thêm số cực nhỏ 0.0000000001 để ko cho x nhận giá trị 0
# # không đồng nhất: p[0.333, 0,33, 0,3333333333]
# p = [0.2, 0.45, 0.35]
# q = [0.31, 0.25, 0.44]
# entropy_p = -sum([p[i]*np.log(p[i]) for i in range(len(p))])
# print(entropy_p)
#
# entropy_q = -sum([q[i]*np.log(q[i]) for i in range(len(q))])
# print(entropy_q)
#
# crossentropy_pq = -sum([p[i]*np.log(q[i]) for i in range(len(p))])
# crossentropy_qp = -sum([q[i]*np.log(p[i]) for i in range(len(q))])
# print(crossentropy_pq, crossentropy_qp)
#
# # crossentropy_pq = np.log(q[i]
#
# # KL Divergence: hệ số phân tán
# KL_Divergence_pq = sum([p[i]*np.log(p[i]/q[i]) for i in range(len(p))])
# print(entropy_q + KL_Divergence_pq)
#
#
# # thông số quan trọng, MSE, likelyhood, crossentropy