近年来恶意软件猖獗，解决其问题刻不容缓。虽用机器学习进行恶意软件分类已较成熟，但它对特征处理要求高，因此我采用深度学习思路，基于Canadian Institute for Cybersecurity提供的近年安卓恶意软件提取特征文件进行改进，初步完成了一个入门级恶意软件分类项目。
☞☞☞AI 智能聊天, 问答助手, AI 智能搜索, 免费无限量使用 DeepSeek R1 模型☜☜☜
![[网络安全篇4] 全连接网络分类实现 - php中文网](https://img.php.cn/upload/article/202507/29/2025072909160630545.jpg)
[网络安全篇4] 全连接网络分类实现
近年来，恶意软件的出现尤为猖獗。所以解决恶意软件的问题也成为了刻不容缓的事情，当然，现在运用机器学习去完成恶意软件分类的手段已经比较的成熟了，但是，机器学习去特征的处理要求是比较高的，这里我采用深度学习的思维去解决。与此同时，Canadian Institute for Cybersecurity这个研究所提供了近几年的安卓恶意软件的提取特征文件，我在此基础上进行改进，初步完成一个较为入门级别的恶意软件分类项目。
0 背景
前面使用了图像的思想去进行分类，发现效果并不好，单纯的采用机器学习的方法都可以达到一个比较不错的结果了，那我们现在就摒弃图像的想法，继续探索正常的分类，其实，一排串的数据是一类，我们其实就可以把他当成文本分类的概念去解决。这样的话，就慢慢的将解决思路转向NLP，然后就可以直接去使用transformer的那些套路完成项目，当然奔着循序渐进的理念，本项目不会涉及transformer，先用全连接进行一个网络搭建。
1 数据集
详情参照CIC研究所的具体发布内容。
https://www.unb.ca/cic/datasets/maldroid-2020.html
![[网络安全篇4] 全连接网络分类实现 - php中文网](https://img.php.cn/upload/article/001/571/248/175375186265674.jpg)
![[网络安全篇4] 全连接网络分类实现 - php中文网](https://img.php.cn/upload/article/001/571/248/175375186324801.jpg)
这里我采用了其中470维度的一个数据进行处理，数据集已经挂载在项目数据集当中。
In [ ]
# 对数据进行一个copy!cp data/data157200/syscallsbinders_frequency_5_Cat.csv work/
2 数据处理
In [ ]
import pandas as pd# 按文件名读取整个文件data_471 = pd.read_csv("work/syscallsbinders_frequency_5_Cat.csv")
data_471.head(2) # 展示前几行数据，通常为展示前5行
ACCESS_PERSONAL_INFO___ ALTER_PHONE_STATE___ ANTI_DEBUG_____ \ 0 1 0 0 1 3 0 0 CREATE_FOLDER_____ CREATE_PROCESS`_____ CREATE_THREAD_____ \ 0 3 0 14 1 6 0 42 DEVICE_ACCESS_____ EXECUTE_____ FS_ACCESS____ FS_ACCESS()____ ... \ 0 2 0 3 0 ... 1 91 0 32 0 ... utimes vfork vibrate vibratePattern wait4 watchRotation \ 0 0 0 0 0 0 0 1 0 0 0 0 0 0 windowGainedFocus write writev Class 0 0 37 10 1 1 2 2838 46 1 [2 rows x 471 columns]
2.2 无效数据过滤
我这里的思维就是，竖向的每一列，如果求和他的值为0，那么他的这一列的数值存在将是毫无意义的。
BJXSHOP网上开店专家
下载
BJXShop网上购物系统是一个高效、稳定、安全的电子商店销售平台，经过近三年市场的考验，在中国网购系统中属领先水平；完善的订单管理、销售统计系统；网站模版可DIY、亦可导入导出；会员、商品种类和价格均实现无限等级；管理员权限可细分；整合了多种在线支付接口；强有力搜索引擎支持... 程序更新：此版本是伴江行官方商业版程序，已经终止销售，现于免费给大家使用。比其以前的免费版功能增加了：1，整合了论坛
换句话来讲，其实这里也是一种特征降维。
前面的图像解决有可以的去将数据控制到一个NxN+1的维度,这里就不再需要这样处理了
我这里只把为全为0的列去除。
（发现并没有全为0的列QAQ）
In [ ]
import pandas as pd# 按文件名读取整个文件data_471 = pd.read_csv("work/syscallsbinders_frequency_5_Cat.csv")
df = pd.DataFrame(data_471)len(df.columns)# 记录每一列求和小于阈值的列threshold = 0de = []# 记录样本数据中的最大值de_max = 0de_va = []for i in range(0,470):
name = df.columns.values[i] if(df[name].sum() < threshold):
de.append(name) if(df[name].max() >= de_max):
de_max = df[name].max()# 删除小于阈值的列for j in de:
df.drop(j,axis=1,inplace=True)
len_new = len(df.columns)
len_new
471
In [ ]
de_max
3697410
In [ ]
# 新数据写入csv 虽然没有变化，新存一下吧，养成一个良好的文件备份习惯df.to_csv("new_date_"+str(len_new)+".csv",index=False,sep=',')
In [ ]
import numpy as npimport pandas as pd
mydata = pd.read_csv("new_date_"+str(len_new)+".csv")
row_list = mydata.values.tolist()# 一个数据展示print(row_list[0])
[1, 0, 0, 3, 0, 14, 2, 0, 3, 0, 3, 0, 0, 0, 4, 0, 9, 0, 7, 2, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 4, 0, 0, 0, 3, 0, 14, 6, 0, 0, 11, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 3, 21, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 4, 0, 0, 109, 14, 24, 0, 7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 3, 0, 0, 3, 0, 0, 0, 0, 0, 20, 0, 0, 0, 0, 2, 0, 14, 0, 2, 0, 0, 170, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 18, 0, 23, 0, 0, 0, 6, 6, 33, 26, 25, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 70, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 5, 61, 3, 0, 0, 61, 0, 133, 0, 0, 0, 22, 8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 28, 0, 0, 0, 0, 2, 0, 0, 0, 0, 4, 96, 12, 0, 0, 0, 0, 0, 0, 0, 0, 29, 0, 0, 3, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 14, 1, 1, 1, 0, 9, 0, 0, 0, 0, 1, 17, 31, 0, 7, 0, 0, 0, 0, 0, 0, 20, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 37, 10, 1]
In [ ]
len(row_list)
11598
In [ ]
# 获取最后一维的数据，也就是分类类别 1-5 这个是一行的分类的展示row_list[0][-1]
1
In [ ]
import numpy as np# 获取类别分布 所有行的最后一列为labelcls_ = np.asarray(mydata.iloc[:-2,-1])print('Labels:', np.unique(cls_))
Labels: [1 2 3 4 5]
In [ ]
_records = {
i: sum([1 for j in cls_ if j==i]) for i in range(1,6)
}print('dist:', _records)
dist: {1: 1253, 2: 2100, 3: 3904, 4: 2546, 5: 1793}
2.3 控制划分后数据类别
防止划分后打乱的数据分类不均衡
In [ ]
import random random.shuffle(row_list)
In [ ]
# 验证一下是不是真的打乱了 因为最原始的数据前面全是1分类，所以下面的输出只要不是全1 就是打乱的了for i in range(4): print(row_list[i][-1])
3 4 3 5
In [ ]
# 划分数据train_rate = 0.8train_list = []
eval_list = []
_cls_records = {i:0 for i in cls_}for data in row_list:
_data_str, _label_int= data[:-1],data[-1:]
label = int(_label_int[0]) # 控制每个类别的比例
if _cls_records[label] < int(_records[label]*train_rate):
train_list.append([_data_str,_label_int]) else:
eval_list.append([_data_str,_label_int])
_cls_records[label] += 1print(len(train_list))print(len(eval_list))print(len(train_list)+len(eval_list))
9275 2323 11598
In [ ]
# 解析数据样本train_data = []
train_label = []for data_label in train_list:
data,label = data_label[0],data_label[1]
train_data.append(data)
train_label.append(label)
val_data = []
val_label = []for data_label in eval_list:
data,label = data_label[0],data_label[1]
val_data.append(data)
val_label.append(label)
In [ ]
print("Train Dataset Length:", len(train_data))print("Eval Dataset Length:", len(val_data))print("Example Sample:", train_data[0], '------', train_label[0])
Train Dataset Length: 9275 Eval Dataset Length: 2323 Example Sample: [2, 0, 0, 3, 0, 25, 4, 0, 9, 0, 2, 0, 0, 2, 2, 0, 14, 0, 7, 4, 0, 0, 0, 4, 0, 8, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 25, 5254, 0, 0, 18, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 70, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 2, 0, 0, 446, 25, 27, 0, 1, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 7, 11, 59, 0, 0, 1, 0, 0, 0, 0, 0, 69, 5, 0, 0, 0, 2, 0, 31, 0, 0, 0, 0, 440, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 10, 10, 10, 0, 0, 0, 36, 0, 30, 0, 0, 0, 0, 3, 177, 10, 40, 0, 0, 0, 0, 1, 0, 3, 0, 0, 0, 0, 0, 0, 91, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 36, 10, 1034, 3, 0, 0, 1029, 0, 285, 0, 0, 0, 14, 7, 0, 0, 0, 0, 0, 0, 0, 0, 0, 40, 0, 0, 0, 0, 8, 0, 0, 0, 0, 0, 1048, 9, 0, 0, 0, 0, 0, 0, 0, 0, 5311, 0, 0, 0, 2, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 26, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 26, 21, 0, 2, 8, 0, 0, 0, 0, 0, 19, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 102, 30] ------ [3]
到这里，数据就处理完了，就可以编写数据处理类和模型网络搭建了。
3 dataset类编写
In [ ]
from paddle.io import Datasetfrom PIL import Imageclass MyDataset(Dataset):
def __init__(self, mode = 'train'):
# 训练样本数量
self.training_data, self.training_label, self.test_data, self.test_label = train_data, train_label, val_data, val_label if mode == 'train':
self.num_samples = len(train_data) else:
self.num_samples = len(val_data)
self.mode = mode def __getitem__(self, idx):
if self.mode == 'train':
data = self.training_data[idx]
label = self.training_label[idx] else:
data = self.test_data[idx]
label = self.test_label[idx]
data = np.array(data).astype('float32')
return data, np.array(label, dtype='int64') def __len__(self):
# 返回样本总数量
return self.num_samples
In [ ]
# 训练的数据提供器train_dataset = MyDataset(mode='train')# 测试的数据提供器eval_dataset = MyDataset(mode='val')# 查看训练和测试数据的大小print('train大小：', train_dataset.__len__())print('eval大小：', eval_dataset.__len__())# 查看图片数据、大小及标签for data, label in train_dataset: print(np.array(data).shape) print(np.array(label).shape) break
train大小： 9275 eval大小： 2323 (470,) (1,)
4 全连接网络搭建
4.1 模型搭建
In [ ]
import paddleimport paddle.nn as nn# 实现MLP模块class Mlp(nn.Layer):
# self代表类的实例自身
def __init__(self):
# 初始化父类中的一些参数
super(Mlp, self).__init__()
self.fc1 = paddle.nn.Linear(in_features=470, out_features=1024)
self.bn1 = paddle.nn.BatchNorm1D(1024)
self.relu1 = paddle.nn.ReLU()
self.fc2 = paddle.nn.Linear(in_features=1024, out_features=512)
self.bn2 = paddle.nn.BatchNorm1D(512)
self.relu2 = paddle.nn.ReLU()
self.fc3 = paddle.nn.Linear(in_features=512, out_features=64)
self.bn3 = paddle.nn.BatchNorm1D(64)
self.relu3 = paddle.nn.ReLU()
self.fc4 = paddle.nn.Linear(in_features=64, out_features=1)
# 网络的前向计算
def forward(self, inputs):
x = self.fc1(inputs)
x = self.bn1(x)
x = self.relu1(x)
x = self.fc2(x)
x = self.bn2(x)
x = self.relu2(x)
x = self.fc3(x)
x = self.bn3(x)
x = self.relu3(x)
x = self.fc4(x) return x
In [ ]
# 测试mlp = Mlp()# 模型结构展示paddle.summary(mlp, (256, 470))
---------------------------------------------------------------------------
Layer (type) Input Shape Output Shape Param #
===========================================================================
Linear-77 [[256, 470]] [256, 1024] 482,304
BatchNorm1D-58 [[256, 1024]] [256, 1024] 4,096
ReLU-58 [[256, 1024]] [256, 1024] 0
Linear-78 [[256, 1024]] [256, 512] 524,800
BatchNorm1D-59 [[256, 512]] [256, 512] 2,048
ReLU-59 [[256, 512]] [256, 512] 0
Linear-79 [[256, 512]] [256, 64] 32,832
BatchNorm1D-60 [[256, 64]] [256, 64] 256
ReLU-60 [[256, 64]] [256, 64] 0
Linear-80 [[256, 64]] [256, 1] 65
===========================================================================
Total params: 1,046,401
Trainable params: 1,040,001
Non-trainable params: 6,400
---------------------------------------------------------------------------
Input size (MB): 0.46
Forward/backward pass size (MB): 9.38
Params size (MB): 3.99
Estimated Total Size (MB): 13.83
---------------------------------------------------------------------------
{'total_params': 1046401, 'trainable_params': 1040001}
4.2 模型训练
In [ ]
import paddlefrom paddle.io import DataLoader
BATCH_SIZE = 256 # 设置batch大小# 训练的数据提供器train_dataloader = DataLoader(
dataset = train_dataset,
batch_size = BATCH_SIZE,
shuffle = True,
drop_last=True
)# 测试的数据提供器eval_dataloader = DataLoader(
dataset = eval_dataset,
batch_size = BATCH_SIZE,
shuffle =False,
drop_last=True
)
In [73]
import os
EPOCH_NUM = 1500 # 设置外层循环次数save_dir = 'output_model/'# 模型封装model = Mlp()
learning_rate = 5e-3opt = paddle.optimizer.SGD(learning_rate=learning_rate, parameters=model.parameters())def train_mlp():
# 定义外层循环
train_loss = []
val_loss = [] for epoch_id in range(EPOCH_NUM): # 开启模型训练模式
model.train() for iter_id, mini_batch in enumerate(train_dataloader()): # 清空梯度变量，以备下一轮计算
opt.clear_grad()
x = np.array(mini_batch[:][0])
y = np.array(mini_batch[:][1]) # 将numpy数据转为飞桨动态图tensor的格式
features = paddle.to_tensor(x,dtype='float32')
y = paddle.to_tensor(y,dtype='float32') # 前向计算
predicts = model(features) # 计算损失
loss = paddle.nn.functional.l1_loss(predicts, label=y)
avg_loss = paddle.mean(loss)
train_loss.append(avg_loss.numpy()) # 反向传播，计算每层参数的梯度值
avg_loss.backward() # 更新参数，根据设置好的学习率迭代一步
opt.step()
# 开启验证模式
model.eval() for iter_id, mini_batch in enumerate(eval_dataloader()):
x = np.array(mini_batch[:][0])
y = np.array(mini_batch[:][1])
features = paddle.to_tensor(x,dtype='float32')
y = paddle.to_tensor(y,dtype='float32')
predicts = model(features)
loss = paddle.nn.functional.l1_loss(predicts, label=y)
avg_loss = paddle.mean(loss)
val_loss.append(avg_loss.numpy())
print(f'Epoch {epoch_id}, train loss is {np.mean(train_loss)}, val loss is {np.mean(val_loss)}') if (epoch_id % 5 == 0): if not os.path.isdir(save_dir):
os.makedirs(save_dir)
save_path = os.path.join(save_dir, '{0}'.format(epoch_id))
paddle.save(model.state_dict(), save_path + '.pdparams')
paddle.save(opt.state_dict(), save_path + '.pdopt') return train_loss,val_loss
In [74]
tloss,vloss = train_mlp()
4.3 训练日志可视化
In [75]
import matplotlib.pyplot as plt
%matplotlib inline
x = [i for i in range(len(tloss))]
y = [i for i in range(len(vloss))]def draw(x,tloss,label):
plt.title(label,fontsize = 24)
plt.xlabel('iter_id',fontsize = 20)
plt.ylabel('loss',fontsize = 20)
plt.plot(x,tloss,color = 'green',label = label)
plt.legend()
plt.grid()
plt.show()
draw(x,tloss,'train loss')
draw(y,vloss,'val loss')
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/__init__.py:107: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working from collections import MutableMapping /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/rcsetup.py:20: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working from collections import Iterable, Mapping /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/colors.py:53: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working from collections import Sized /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2349: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working if isinstance(obj, collections.Iterator): /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2366: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working return list(data) if isinstance(data, collections.MappingView) else data
5 模型保存及简单验证
5.1 模型保存
In [76]
paddle.save(model.state_dict(), 'out/final.pdparams') # 保存模型
5.2 模型验证
其实上面的可视化图已经是对结果的一个很好的验证了，这里我们通过实际的数据的展示，来看一下效果。
In [77]
print(val_data[10])print(val_label[10][0])
[0, 0, 0, 3, 0, 9, 2, 0, 3, 0, 3, 0, 0, 0, 3, 0, 8, 0, 6, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 9, 6, 0, 0, 6, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 20, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 3, 0, 0, 109, 9, 19, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 31, 0, 0, 1, 0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 2, 0, 8, 0, 1, 0, 0, 1184, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 27, 0, 30, 0, 0, 0, 0, 3, 23, 9, 46, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 103, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 4, 47, 3, 0, 0, 35, 0, 183, 0, 0, 0, 5, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 25, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 91, 1, 0, 0, 0, 0, 0, 0, 0, 0, 24, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 6, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 10, 1, 0, 1, 0, 0, 0, 0, 0, 0, 29, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 65, 3] 2
In [80]
import paddle.nn.functional as F
label_txt = ['','Adware','Banking','SMS malware','Riskware','Benign']
model = Mlp()
para_dict = paddle.load('out/final.pdparams')
model.load_dict(para_dict)
model.eval()
x = paddle.reshape(paddle.to_tensor(val_data[10],dtype='float32'),[-1,470])
predicts = model(x)
pre = paddle.round(predicts)print("实际标签为:",label_txt[val_label[10][0]])print("预测标签为:",label_txt[pre[0]])
实际标签为: Banking 预测标签为: Banking
