资源简介
基于Tensorflow用CNN(卷积神经网络)处理kdd99数据集,代码包括预处理代码和分类代码,准确率99.6%以上,并且快速收敛至最优值。
(Based on Tensorflow (convolutional neural network) processing KDD99 data set based on CNN, the code includes preprocessing code and classification code, the accuracy rate is more than 99.6%, and quickly converge to the optimal value.)
代码片段和文件信息
#/usr/bin/python2.7
#coding:utf-8
from __future__ import print_function
import tensorflow as tf
import randomcsv
def next_batch(feature_listlabel_listsize):
feature_batch_temp=[]
label_batch_temp=[]
f_list = random.sample(range(len(feature_list)) size)
for i in f_list:
feature_batch_temp.append(feature_list[i])
for i in f_list:
label_batch_temp.append(label_list[i])
return feature_batch_templabel_batch_temp
def weight_variable(shapelayer_name):
#定义一个shape形状的weights张量
with tf.name_scope(layer_name + ‘_Weights‘):
Weights = tf.Variable(tf.truncated_normal(shape stddev=0.1)name=‘W‘)
tf.histogram_summary(layer_name + ‘_Weights‘ Weights)
return Weights
def bias_variable(shapelayer_name):
#定义一个shape形状的bias张量
with tf.name_scope(layer_name + ‘_biases‘):
biases = tf.Variable(tf.constant(0.1 shape=shape)name=‘b‘)
tf.histogram_summary(layer_name + ‘_biases‘ biases)
return biases
def conv2d(x Wlayer_name):
#卷积计算函数
# stride [1 x步长 y步长 1]
# padding:SAME/FULL/VALID(边距处理方式)
with tf.name_scope(layer_name + ‘_h_conv2d‘):
h_conv2d = tf.nn.conv2d(x W strides=[1 1 1 1] padding=‘SAME‘)
return h_conv2d
def max_pool_2x2(xlayer_name):
# max池化函数
# ksize [1 x边长 y边长1] 池化窗口大小
# stride [1 x步长 y步长 1]
# padding:SAME/FULL/VALID(边距处理方式)
with tf.name_scope(layer_name + ‘_h_pool‘):
h_pool = tf.nn.max_pool(x ksize=[1221] strides=[1221] padding=‘SAME‘)
return h_pool
def load_data():
global feature
global label
global feature_full
global label_full
feature=[]
label=[]
feature_full=[]
label_full=[]
file_path =‘/home/peter/Desktop/pycharm/ids-kdd99/kddcup.data_10_percent_corrected_handled2.cvs‘
with (open(file_path‘r‘)) as data_from:
csv_reader=csv.reader(data_from)
for i in csv_reader:
# print i
label_list=[0]*23
feature.append(i[:36])
label_list[int(i[41])]=1
label.append(label_list)
# print label
# print feature
file_path_full =‘/home/peter/Desktop/pycharm/ids-kdd99/kddcup.data.corrected_handled2.cvs‘
with (open(file_path_full‘r‘)) as data_from_full:
csv_reader_full=csv.reader(data_from_full)
for j in csv_reader_full:
# print i
label_list_full=[0]*23
feature_full.append(j[:36])
label_list_full[int(j[41])]=1
label_full.append(label_list_full)
if __name__ == ‘__main__‘:
global feature
global label
global feature_full
global label_full
# load数据
load_data()
feature_test = feature
feature_train =feature_full
label_test = label
label_test_full = label_full
# 定义用以输入的palceholder
with tf.name_scope(‘inputs‘):
xs = tf.placeholder(tf.float32 [None 36]name=‘pic_data‘) # 6x6
ys = tf.placeholder(tf.float32 [None 23]
属性 大小 日期 时间 名称
----------- --------- ---------- ----- ----
目录 0 2017-06-08 20:49 ids-kdd99\
目录 0 2017-06-08 20:49 ids-kdd99\.idea\
文件 398 2016-12-27 15:54 ids-kdd99\.idea\ids-kdd99.iml
文件 682 2016-12-27 15:53 ids-kdd99\.idea\misc.xm
文件 270 2016-12-27 15:53 ids-kdd99\.idea\modules.xm
文件 42708 2016-12-29 10:01 ids-kdd99\.idea\workspace.xm
文件 6944 2016-12-29 16:58 ids-kdd99\cnn_main.py
文件 2977 2016-12-29 16:55 ids-kdd99\handle2.py
文件 18115902 2016-12-29 09:58 ids-kdd99\kddcup.data.gz
文件 2144903 2016-12-28 16:45 ids-kdd99\kddcup.data_10_percent.gz
文件 4659 2016-12-29 17:00 ids-kdd99\main.py
文件 6944 2017-02-27 15:12 ids-kdd99\mian_cnn.py
目录 0 2017-06-08 20:49 ids-kdd99\multi_logs\
文件 53246 2016-12-29 11:38 ids-kdd99\multi_logs\events.out.tfevents.1482980284.zjx-24000635
目录 0 2017-06-08 20:49 ids-kdd99\multi_logs\test\
文件 155823 2016-12-29 11:38 ids-kdd99\multi_logs\test\events.out.tfevents.1482980284.zjx-24000635
目录 0 2017-06-08 20:49 ids-kdd99\multi_logs\train\
文件 155823 2016-12-29 11:38 ids-kdd99\multi_logs\train\events.out.tfevents.1482980284.zjx-24000635
文件 328 2016-12-29 17:11 ids-kdd99\readMe.txt.txt
相关资源
- 入侵检测试验
- 卷积神经网络基础及发展现状
- CNN卷积神经网络应用于人脸识别带详
- 卷积神经网络文字识别
- CIC-IDS2018数据集
- [Snort轻量级入侵检测系统全攻略].陈伟
- 卷积神经网络+网络优化算法+目标检测
- 魏秀参:解析卷积神经网络
- 吴恩达老师深度学习第四课卷积神经
- 网络入侵检测系统kddcup1999数据集
- kdd 99 全部数据集
- eyeriss项目组的深度学习加速器的总结
- NSL-KDD数据集+KDD数据集
- 卷积神经网络车牌识别164048
- 卷积神经网络车牌识别
- 现流行的AlexNetVGGNetGoogleNetSENetResNet等
- 深度学习卷积神经网络代码
- 联想网御入侵检测系统用户手册
- 手写数字识别MNIST数据集及卷积神经网
- 卷积神经网络CNNs-(从AlexNet开始)
- 网络入侵检测系统的设计与实现绝版
- CS231 卷积神经网络(中文版带书签)
- 虚拟蜜罐从僵尸网络追踪到入侵检测
- CNN卷积神经网络识别手写汉字MNIST数据
- 卷积神经网络CNN进行图像分类
- 卷积神经网络实现情感分类
- 吴恩达卷积神经网络课件与笔记
- 卷积神经网络图像分类和检测必看论
- 基于CNN和SVM的猫狗识别
- 卷积神经网络实现车牌识别.zip
评论
共有 条评论