# 导入模块
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
print(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 = [-1, -1, -1, -2, -3, -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,根据总位移得到的行为轨迹
tracks = get_tracks(distance)
print(tracks) # 7.按照人行动轨迹先正向滑动,后反向滑动
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,自信的表现是疯狂加速
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(0.3)
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, 这一步可以极大的提高成功率
time.sleep(0.1)
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='aaaaa',password='aaaaa')