爬虫-破解验证码

一 介绍

    一些网站会在正常的账号密码认证之外加一些验证码,以此来明确地区分人/机行为,从一定程度上达到反爬的效果,对于简单的校验码Tesserocr就可以搞定,如下

    但一些网站加入了滑动验证码,最典型的要属于极验滑动认证了,极验官网:http://www.geetest.com/,下图是极验的登录界面

    现在极验验证码已经更新到了 3.0 版本,截至 2017 年 7 月全球已有十六万家企业正在使用极验,每天服务响应超过四亿次,广泛应用于直播视频、金融服务、电子商务、游戏娱乐、政府企业等各大类型网站

对于这类验证,如果我们直接模拟表单请求,繁琐的认证参数与认证流程会让你蛋碎一地,我们可以用selenium驱动浏览器来解决这个问题,大致分为以下几个步骤

复制代码
#步骤一:点击按钮,弹出没有缺口的图片

#步骤二:获取步骤一的图片

#步骤三:点击滑动按钮,弹出带缺口的图片

#步骤四:获取带缺口的图片

#步骤五:对比两张图片的所有RBG像素点,得到不一样像素点的x值,即要移动的距离

#步骤六:模拟人的行为习惯(先匀加速拖动后匀减速拖动),把需要拖动的总距离分成一段一段小的轨迹

#步骤七:按照轨迹拖动,完全验证

#步骤八:完成登录
复制代码

二 实现

复制代码
#安装:selenium+chrome/phantomjs

#安装:Pillow
Pillow:基于PIL,处理python 3.x的图形图像库.因为PIL只能处理到python 2.x,而这个模块能处理Python3.x,目前用它做图形的很多.
http://www.cnblogs.com/apexchu/p/4231041.html

C:\Users\Administrator>pip3 install pillow
C:\Users\Administrator>python3
Python 3.6.1 (v3.6.1:69c0db5, Mar 21 2017, 18:41:36) [MSC v.1900 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> from PIL import Image
>>>
复制代码
from selenium import webdriver
from selenium.webdriver import ActionChains
from selenium.webdriver.common.by import By
from selenium.webdriver.common.keys import Keys
from selenium.webdriver.support import expected_conditions as EC
from selenium.webdriver.support.wait import WebDriverWait
from PIL import Image
import time

def get_snap():
'''
对整个网页截图,保存成图片,然后用PIL.Image拿到图片对象
:return: 图片对象
'''
driver.save_screenshot('snap.png')
page_snap_obj=Image.open('snap.png')
return page_snap_obj

def get_image():
'''
从网页的网站截图中,截取验证码图片
:return: 验证码图片
'''
img=wait.until(EC.presence_of_element_located((By.CLASS_NAME,'geetest_canvas_img')))
time.sleep(2) #保证图片刷新出来
localtion=img.location
size=img.size

top=localtion['y']
bottom=localtion['y']+size['height']
left=localtion['x']
right=localtion['x']+size['width']

page_snap_obj=get_snap()
crop_imag_obj=page_snap_obj.crop((left,top,right,bottom))
return crop_imag_obj


def get_distance(image1,image2):
'''
拿到滑动验证码需要移动的距离
:param image1:没有缺口的图片对象
:param image2:带缺口的图片对象
:return:需要移动的距离
'''
threshold=60
left=57
for i in range(left,image1.size[0]):
for j in range(image1.size[1]):
rgb1=image1.load()[i,j]
rgb2=image2.load()[i,j]
res1=abs(rgb1[0]-rgb2[0])
res2=abs(rgb1[1]-rgb2[1])
res3=abs(rgb1[2]-rgb2[2])
if not (res1 < threshold and res2 < threshold and res3 < threshold):
return i-7 #经过测试,误差为大概为7
return i-7 #经过测试,误差为大概为7


def get_tracks(distance):
'''
拿到移动轨迹,模仿人的滑动行为,先匀加速后匀减速
匀变速运动基本公式:
①v=v0+at
②s=v0t+½at²
③v²-v0²=2as

:param distance: 需要移动的距离
:return: 存放每0.3秒移动的距离
'''
#初速度
v=0
#单位时间为0.2s来统计轨迹,轨迹即0.2内的位移
t=0.3
#位移/轨迹列表,列表内的一个元素代表0.2s的位移
tracks=[]
#当前的位移
current=0
#到达mid值开始减速
mid=distance*4/5

while current < distance:
if current < mid:
# 加速度越小,单位时间的位移越小,模拟的轨迹就越多越详细
a= 2
else:
a=-3

#初速度
v0=v
#0.2秒时间内的位移
s=v0*t+0.5*a*(t**2)
#当前的位置
current+=s
#添加到轨迹列表
 tracks.append(round(s))

#速度已经达到v,该速度作为下次的初速度
v=v0+a*t
return tracks


try:
driver=webdriver.Chrome()
driver.get('https://account.geetest.com/login')
wait=WebDriverWait(driver,10)

#步骤一:先点击按钮,弹出没有缺口的图片
button=wait.until(EC.presence_of_element_located((By.CLASS_NAME,'geetest_radar_tip')))
button.click()

#步骤二:拿到没有缺口的图片
image1=get_image()

#步骤三:点击拖动按钮,弹出有缺口的图片
button=wait.until(EC.presence_of_element_located((By.CLASS_NAME,'geetest_slider_button')))
button.click()

#步骤四:拿到有缺口的图片
image2=get_image()

# print(image1,image1.size)
# print(image2,image2.size)

#步骤五:对比两张图片的所有RBG像素点,得到不一样像素点的x值,即要移动的距离
distance=get_distance(image1,image2)

#步骤六:模拟人的行为习惯(先匀加速拖动后匀减速拖动),把需要拖动的总距离分成一段一段小的轨迹
tracks=get_tracks(distance)
print(tracks)
print(image1.size)
print(distance,sum(tracks))


#步骤七:按照轨迹拖动,完全验证
button=wait.until(EC.presence_of_element_located((By.CLASS_NAME,'geetest_slider_button')))
ActionChains(driver).click_and_hold(button).perform()
for track in tracks:
ActionChains(driver).move_by_offset(xoffset=track,yoffset=0).perform()
else:
ActionChains(driver).move_by_offset(xoffset=3,yoffset=0).perform() #先移过一点
ActionChains(driver).move_by_offset(xoffset=-3,yoffset=0).perform() #再退回来,是不是更像人了

time.sleep(0.5) #0.5秒后释放鼠标
 ActionChains(driver).release().perform()


#步骤八:完成登录
input_email=driver.find_element_by_id('email')
input_password=driver.find_element_by_id('password')
button=wait.until(EC.element_to_be_clickable((By.CLASS_NAME,'login-btn')))

input_email.send_keys('18611453110@163.com')
input_password.send_keys('linhaifeng123')
# button.send_keys(Keys.ENTER)
 button.click()

import time
time.sleep(200)
finally:
driver.close()
View Code

案例:

from selenium import webdriver
from selenium.webdriver import ActionChains
from selenium.webdriver.common.by import By
from selenium.webdriver.common.keys import Keys
from selenium.webdriver.support import expected_conditions as EC
from selenium.webdriver.support.wait import WebDriverWait
from PIL import Image
import time

def get_snap():
driver.save_screenshot('full_snap.png')
page_snap_obj=Image.open('full_snap.png')
return page_snap_obj

def get_image():
img=driver.find_element_by_class_name('geetest_canvas_img')
time.sleep(2)
location=img.location
size=img.size

left=location['x']
top=location['y']
right=left+size['width']
bottom=top+size['height']

page_snap_obj=get_snap()
image_obj=page_snap_obj.crop((left,top,right,bottom))
# image_obj.show()
return image_obj

def get_distance(image1,image2):
start=57
threhold=60

for i in range(start,image1.size[0]):
for j in range(image1.size[1]):
rgb1=image1.load()[i,j]
rgb2=image2.load()[i,j]
res1=abs(rgb1[0]-rgb2[0])
res2=abs(rgb1[1]-rgb2[1])
res3=abs(rgb1[2]-rgb2[2])
# print(res1,res2,res3)
if not (res1 < threhold and res2 < threhold and res3 < threhold):
return i-7
return i-7

def get_tracks(distance):
distance+=20 #先滑过一点,最后再反着滑动回来
v=0
t=0.2
forward_tracks=[]

current=0
mid=distance*3/5
while current < distance:
if current < mid:
a=2
else:
a=-3

s=v*t+0.5*a*(t**2)
v=v+a*t
current+=s
forward_tracks.append(round(s))

#反着滑动到准确位置
back_tracks=[-3,-3,-2,-2,-2,-2,-2,-1,-1,-1] #总共等于-20

return {'forward_tracks':forward_tracks,'back_tracks':back_tracks}

try:
# 1、输入账号密码回车
driver = webdriver.Chrome()
driver.implicitly_wait(3)
driver.get('https://passport.cnblogs.com/user/signin')

username = driver.find_element_by_id('input1')
pwd = driver.find_element_by_id('input2')
signin = driver.find_element_by_id('signin')

username.send_keys('linhaifeng')
pwd.send_keys('xxxxx')
signin.click()

# 2、点击按钮,得到没有缺口的图片
button = driver.find_element_by_class_name('geetest_radar_tip')
button.click()

# 3、获取没有缺口的图片
image1 = get_image()

# 4、点击滑动按钮,得到有缺口的图片
button = driver.find_element_by_class_name('geetest_slider_button')
button.click()

# 5、获取有缺口的图片
image2 = get_image()

# 6、对比两种图片的像素点,找出位移
distance = get_distance(image1, image2)

# 7、模拟人的行为习惯,根据总位移得到行为轨迹
tracks = get_tracks(distance)
print(tracks)

# 8、按照行动轨迹先正向滑动,后反滑动
button = driver.find_element_by_class_name('geetest_slider_button')
ActionChains(driver).click_and_hold(button).perform()

# 正常人类总是自信满满地开始正向滑动,自信地表现是疯狂加速
for track in tracks['forward_tracks']:
ActionChains(driver).move_by_offset(xoffset=track, yoffset=0).perform()

# 结果傻逼了,正常的人类停顿了一下,回过神来发现,卧槽,滑过了,然后开始反向滑动
time.sleep(0.5)
for back_track in tracks['back_tracks']:
ActionChains(driver).move_by_offset(xoffset=back_track, yoffset=0).perform()

# 小范围震荡一下,进一步迷惑极验后台,这一步可以极大地提高成功率
ActionChains(driver).move_by_offset(xoffset=-3, yoffset=0).perform()
ActionChains(driver).move_by_offset(xoffset=3, yoffset=0).perform()

# 成功后,骚包人类总喜欢默默地欣赏一下自己拼图的成果,然后恋恋不舍地松开那只脏手
time.sleep(0.5)
ActionChains(driver).release().perform()

time.sleep(10) # 睡时间长一点,确定登录成功
finally:
driver.close()
破解博客园后台登录
from selenium import webdriver
from selenium.webdriver import ActionChains
from selenium.webdriver.common.by import By
from selenium.webdriver.common.keys import Keys
from selenium.webdriver.support import expected_conditions as EC
from selenium.webdriver.support.wait import WebDriverWait
from PIL import Image
import time

def get_snap(driver):
driver.save_screenshot('full_snap.png')
page_snap_obj=Image.open('full_snap.png')
return page_snap_obj

def get_image(driver):
img=driver.find_element_by_class_name('geetest_canvas_img')
time.sleep(2)
location=img.location
size=img.size

left=location['x']
top=location['y']
right=left+size['width']
bottom=top+size['height']

page_snap_obj=get_snap(driver)
image_obj=page_snap_obj.crop((left,top,right,bottom))
# image_obj.show()
return image_obj

def get_distance(image1,image2):
start=57
threhold=60

for i in range(start,image1.size[0]):
for j in range(image1.size[1]):
rgb1=image1.load()[i,j]
rgb2=image2.load()[i,j]
res1=abs(rgb1[0]-rgb2[0])
res2=abs(rgb1[1]-rgb2[1])
res3=abs(rgb1[2]-rgb2[2])
# print(res1,res2,res3)
if not (res1 < threhold and res2 < threhold and res3 < threhold):
return i-7
return i-7

def get_tracks(distance):
distance+=20 #先滑过一点,最后再反着滑动回来
v=0
t=0.2
forward_tracks=[]

current=0
mid=distance*3/5
while current < distance:
if current < mid:
a=2
else:
a=-3

s=v*t+0.5*a*(t**2)
v=v+a*t
current+=s
forward_tracks.append(round(s))

#反着滑动到准确位置
back_tracks=[-3,-3,-2,-2,-2,-2,-2,-1,-1,-1] #总共等于-20

return {'forward_tracks':forward_tracks,'back_tracks':back_tracks}

def crack(driver): #破解滑动认证
# 1、点击按钮,得到没有缺口的图片
button = driver.find_element_by_class_name('geetest_radar_tip')
button.click()

# 2、获取没有缺口的图片
image1 = get_image(driver)

# 3、点击滑动按钮,得到有缺口的图片
button = driver.find_element_by_class_name('geetest_slider_button')
button.click()

# 4、获取有缺口的图片
image2 = get_image(driver)

# 5、对比两种图片的像素点,找出位移
distance = get_distance(image1, image2)

# 6、模拟人的行为习惯,根据总位移得到行为轨迹
tracks = get_tracks(distance)
print(tracks)

# 7、按照行动轨迹先正向滑动,后反滑动
button = driver.find_element_by_class_name('geetest_slider_button')
ActionChains(driver).click_and_hold(button).perform()

# 正常人类总是自信满满地开始正向滑动,自信地表现是疯狂加速
for track in tracks['forward_tracks']:
ActionChains(driver).move_by_offset(xoffset=track, yoffset=0).perform()

# 结果傻逼了,正常的人类停顿了一下,回过神来发现,卧槽,滑过了,然后开始反向滑动
time.sleep(0.5)
for back_track in tracks['back_tracks']:
ActionChains(driver).move_by_offset(xoffset=back_track, yoffset=0).perform()

# 小范围震荡一下,进一步迷惑极验后台,这一步可以极大地提高成功率
ActionChains(driver).move_by_offset(xoffset=-3, yoffset=0).perform()
ActionChains(driver).move_by_offset(xoffset=3, yoffset=0).perform()

# 成功后,骚包人类总喜欢默默地欣赏一下自己拼图的成果,然后恋恋不舍地松开那只脏手
time.sleep(0.5)
ActionChains(driver).release().perform()

def login_cnblogs(username,password):
driver = webdriver.Chrome()
try:
# 1、输入账号密码回车
driver.implicitly_wait(3)
driver.get('https://passport.cnblogs.com/user/signin')

input_username = driver.find_element_by_id('input1')
input_pwd = driver.find_element_by_id('input2')
signin = driver.find_element_by_id('signin')

input_username.send_keys(username)
input_pwd.send_keys(password)
signin.click()

# 2、破解滑动认证
 crack(driver)

time.sleep(10) # 睡时间长一点,确定登录成功
finally:
driver.close()

if __name__ == '__main__':
login_cnblogs(username='linhaifeng',password='xxxx')
修订版

三 说明

  面对简单的滑动验证码,极验其实是有更复杂版本的,如下所示

机器识别难度高了,大部分屌丝码农搞不定了。然而人类也蒙蔽了,易用性降到极低。

使用了上述验证的网站常常会在用户一片怨声载道中,又将其恢复成易于破解的滑动验证。

验证过程,是个破解难度、用户体验之间的一个平衡点。体验越好的,破解也越容易。
嘲讽验证码无效,破解简单,是很 LOW 的行为。

网站方、验证码平台方,知道你能破解,你牛 B。。。更难的验证码他们也有,只是这会严重降低体验,他们不用而已。

 

 

posted @ 2018-01-18 22:42  forjie  阅读(2765)  评论(1编辑  收藏  举报