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
#1、输入账号、密码,然后点击登陆 #2、点击按钮,弹出没有缺口的图 #3、针对没有缺口的图片进行截图 #4、点击滑动按钮,弹出有缺口的图 #5、针对有缺口的图片进行截图 #6、对比两张图片,找出缺口,即滑动的位移 #7、按照人的行为行为习惯,把总位移切成一段段小的位移 #8、按照位移移动
#9、完成登录
二、实现
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:UsersAdministrator>pip3 install pillow C:UsersAdministrator>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 >>>
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
![](https://images.cnblogs.com/OutliningIndicators/ExpandedBlockStart.gif)
1 from selenium import webdriver 2 from selenium.webdriver import ActionChains 3 from selenium.webdriver.common.by import By 4 from selenium.webdriver.common.keys import Keys 5 from selenium.webdriver.support import expected_conditions as EC 6 from selenium.webdriver.support.wait import WebDriverWait 7 from PIL import Image 8 import time 9 10 def get_snap(): 11 ''' 12 对整个网页截图,保存成图片,然后用PIL.Image拿到图片对象 13 :return: 图片对象 14 ''' 15 driver.save_screenshot('snap.png') 16 page_snap_obj=Image.open('snap.png') 17 return page_snap_obj 18 19 def get_image(): 20 ''' 21 从网页的网站截图中,截取验证码图片 22 :return: 验证码图片 23 ''' 24 img=wait.until(EC.presence_of_element_located((By.CLASS_NAME,'geetest_canvas_img'))) 25 time.sleep(2) #保证图片刷新出来 26 localtion=img.location 27 size=img.size 28 29 top=localtion['y'] 30 bottom=localtion['y']+size['height'] 31 left=localtion['x'] 32 right=localtion['x']+size['width'] 33 34 page_snap_obj=get_snap() 35 crop_imag_obj=page_snap_obj.crop((left,top,right,bottom)) 36 return crop_imag_obj 37 38 39 def get_distance(image1,image2): 40 ''' 41 拿到滑动验证码需要移动的距离 42 :param image1:没有缺口的图片对象 43 :param image2:带缺口的图片对象 44 :return:需要移动的距离 45 ''' 46 threshold=60 47 left=57 48 for i in range(left,image1.size[0]): 49 for j in range(image1.size[1]): 50 rgb1=image1.load()[i,j] 51 rgb2=image2.load()[i,j] 52 res1=abs(rgb1[0]-rgb2[0]) 53 res2=abs(rgb1[1]-rgb2[1]) 54 res3=abs(rgb1[2]-rgb2[2]) 55 if not (res1 < threshold and res2 < threshold and res3 < threshold): 56 return i-7 #经过测试,误差为大概为7 57 return i-7 #经过测试,误差为大概为7 58 59 60 def get_tracks(distance): 61 ''' 62 拿到移动轨迹,模仿人的滑动行为,先匀加速后匀减速 63 匀变速运动基本公式: 64 ①v=v0+at 65 ②s=v0t+½at² 66 ③v²-v0²=2as 67 68 :param distance: 需要移动的距离 69 :return: 存放每0.3秒移动的距离 70 ''' 71 #初速度 72 v=0 73 #单位时间为0.2s来统计轨迹,轨迹即0.2内的位移 74 t=0.3 75 #位移/轨迹列表,列表内的一个元素代表0.2s的位移 76 tracks=[] 77 #当前的位移 78 current=0 79 #到达mid值开始减速 80 mid=distance*4/5 81 82 while current < distance: 83 if current < mid: 84 # 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 85 a= 2 86 else: 87 a=-3 88 89 #初速度 90 v0=v 91 #0.2秒时间内的位移 92 s=v0*t+0.5*a*(t**2) 93 #当前的位置 94 current+=s 95 #添加到轨迹列表 96 tracks.append(round(s)) 97 98 #速度已经达到v,该速度作为下次的初速度 99 v=v0+a*t 100 return tracks 101 102 103 try: 104 driver=webdriver.Chrome() 105 driver.get('https://account.geetest.com/login') 106 wait=WebDriverWait(driver,10) 107 108 #步骤一:先点击按钮,弹出没有缺口的图片 109 button=wait.until(EC.presence_of_element_located((By.CLASS_NAME,'geetest_radar_tip'))) 110 button.click() 111 112 #步骤二:拿到没有缺口的图片 113 image1=get_image() 114 115 #步骤三:点击拖动按钮,弹出有缺口的图片 116 button=wait.until(EC.presence_of_element_located((By.CLASS_NAME,'geetest_slider_button'))) 117 button.click() 118 119 #步骤四:拿到有缺口的图片 120 image2=get_image() 121 122 # print(image1,image1.size) 123 # print(image2,image2.size) 124 125 #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 126 distance=get_distance(image1,image2) 127 128 #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 129 tracks=get_tracks(distance) 130 print(tracks) 131 print(image1.size) 132 print(distance,sum(tracks)) 133 134 135 #步骤七:按照轨迹拖动,完全验证 136 button=wait.until(EC.presence_of_element_located((By.CLASS_NAME,'geetest_slider_button'))) 137 ActionChains(driver).click_and_hold(button).perform() 138 for track in tracks: 139 ActionChains(driver).move_by_offset(xoffset=track,yoffset=0).perform() 140 else: 141 ActionChains(driver).move_by_offset(xoffset=3,yoffset=0).perform() #先移过一点 142 ActionChains(driver).move_by_offset(xoffset=-3,yoffset=0).perform() #再退回来,是不是更像人了 143 144 time.sleep(0.5) #0.5秒后释放鼠标 145 ActionChains(driver).release().perform() 146 147 148 #步骤八:完成登录 149 input_email=driver.find_element_by_id('email') 150 input_password=driver.find_element_by_id('password') 151 button=wait.until(EC.element_to_be_clickable((By.CLASS_NAME,'login-btn'))) 152 153 input_email.send_keys('18611453110@163.com') 154 input_password.send_keys('linhaifeng123') 155 # button.send_keys(Keys.ENTER) 156 button.click() 157 158 import time 159 time.sleep(200) 160 finally: 161 driver.close()
案例:
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
![](https://images.cnblogs.com/OutliningIndicators/ExpandedBlockStart.gif)
1 from selenium import webdriver 2 from selenium.webdriver import ActionChains 3 from selenium.webdriver.common.by import By 4 from selenium.webdriver.common.keys import Keys 5 from selenium.webdriver.support import expected_conditions as EC 6 from selenium.webdriver.support.wait import WebDriverWait 7 from PIL import Image 8 import time 9 10 def get_snap(): 11 driver.save_screenshot('full_snap.png') 12 page_snap_obj=Image.open('full_snap.png') 13 return page_snap_obj 14 15 def get_image(): 16 img=driver.find_element_by_class_name('geetest_canvas_img') 17 time.sleep(2) 18 location=img.location 19 size=img.size 20 21 left=location['x'] 22 top=location['y'] 23 right=left+size['width'] 24 bottom=top+size['height'] 25 26 page_snap_obj=get_snap() 27 image_obj=page_snap_obj.crop((left,top,right,bottom)) 28 # image_obj.show() 29 return image_obj 30 31 def get_distance(image1,image2): 32 start=57 33 threhold=60 34 35 for i in range(start,image1.size[0]): 36 for j in range(image1.size[1]): 37 rgb1=image1.load()[i,j] 38 rgb2=image2.load()[i,j] 39 res1=abs(rgb1[0]-rgb2[0]) 40 res2=abs(rgb1[1]-rgb2[1]) 41 res3=abs(rgb1[2]-rgb2[2]) 42 # print(res1,res2,res3) 43 if not (res1 < threhold and res2 < threhold and res3 < threhold): 44 return i-7 45 return i-7 46 47 def get_tracks(distance): 48 distance+=20 #先滑过一点,最后再反着滑动回来 49 v=0 50 t=0.2 51 forward_tracks=[] 52 53 current=0 54 mid=distance*3/5 55 while current < distance: 56 if current < mid: 57 a=2 58 else: 59 a=-3 60 61 s=v*t+0.5*a*(t**2) 62 v=v+a*t 63 current+=s 64 forward_tracks.append(round(s)) 65 66 #反着滑动到准确位置 67 back_tracks=[-3,-3,-2,-2,-2,-2,-2,-1,-1,-1] #总共等于-20 68 69 return {'forward_tracks':forward_tracks,'back_tracks':back_tracks} 70 71 try: 72 # 1、输入账号密码回车 73 driver = webdriver.Chrome() 74 driver.implicitly_wait(3) 75 driver.get('https://passport.cnblogs.com/user/signin') 76 77 username = driver.find_element_by_id('input1') 78 pwd = driver.find_element_by_id('input2') 79 signin = driver.find_element_by_id('signin') 80 81 username.send_keys('linhaifeng') 82 pwd.send_keys('xxxxx') 83 signin.click() 84 85 # 2、点击按钮,得到没有缺口的图片 86 button = driver.find_element_by_class_name('geetest_radar_tip') 87 button.click() 88 89 # 3、获取没有缺口的图片 90 image1 = get_image() 91 92 # 4、点击滑动按钮,得到有缺口的图片 93 button = driver.find_element_by_class_name('geetest_slider_button') 94 button.click() 95 96 # 5、获取有缺口的图片 97 image2 = get_image() 98 99 # 6、对比两种图片的像素点,找出位移 100 distance = get_distance(image1, image2) 101 102 # 7、模拟人的行为习惯,根据总位移得到行为轨迹 103 tracks = get_tracks(distance) 104 print(tracks) 105 106 # 8、按照行动轨迹先正向滑动,后反滑动 107 button = driver.find_element_by_class_name('geetest_slider_button') 108 ActionChains(driver).click_and_hold(button).perform() 109 110 # 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 111 for track in tracks['forward_tracks']: 112 ActionChains(driver).move_by_offset(xoffset=track, yoffset=0).perform() 113 114 # 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 115 time.sleep(0.5) 116 for back_track in tracks['back_tracks']: 117 ActionChains(driver).move_by_offset(xoffset=back_track, yoffset=0).perform() 118 119 # 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 120 ActionChains(driver).move_by_offset(xoffset=-3, yoffset=0).perform() 121 ActionChains(driver).move_by_offset(xoffset=3, yoffset=0).perform() 122 123 # 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 124 time.sleep(0.5) 125 ActionChains(driver).release().perform() 126 127 time.sleep(10) # 睡时间长一点,确定登录成功 128 finally: 129 driver.close()
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
![](https://images.cnblogs.com/OutliningIndicators/ExpandedBlockStart.gif)
1 from selenium import webdriver 2 from selenium.webdriver import ActionChains 3 from selenium.webdriver.common.by import By 4 from selenium.webdriver.common.keys import Keys 5 from selenium.webdriver.support import expected_conditions as EC 6 from selenium.webdriver.support.wait import WebDriverWait 7 from PIL import Image 8 import time 9 10 def get_snap(driver): 11 driver.save_screenshot('full_snap.png') 12 page_snap_obj=Image.open('full_snap.png') 13 return page_snap_obj 14 15 def get_image(driver): 16 img=driver.find_element_by_class_name('geetest_canvas_img') 17 time.sleep(2) 18 location=img.location 19 size=img.size 20 21 left=location['x'] 22 top=location['y'] 23 right=left+size['width'] 24 bottom=top+size['height'] 25 26 page_snap_obj=get_snap(driver) 27 image_obj=page_snap_obj.crop((left,top,right,bottom)) 28 # image_obj.show() 29 return image_obj 30 31 def get_distance(image1,image2): 32 start=57 33 threhold=60 34 35 for i in range(start,image1.size[0]): 36 for j in range(image1.size[1]): 37 rgb1=image1.load()[i,j] 38 rgb2=image2.load()[i,j] 39 res1=abs(rgb1[0]-rgb2[0]) 40 res2=abs(rgb1[1]-rgb2[1]) 41 res3=abs(rgb1[2]-rgb2[2]) 42 # print(res1,res2,res3) 43 if not (res1 < threhold and res2 < threhold and res3 < threhold): 44 return i-7 45 return i-7 46 47 def get_tracks(distance): 48 distance+=20 #先滑过一点,最后再反着滑动回来 49 v=0 50 t=0.2 51 forward_tracks=[] 52 53 current=0 54 mid=distance*3/5 55 while current < distance: 56 if current < mid: 57 a=2 58 else: 59 a=-3 60 61 s=v*t+0.5*a*(t**2) 62 v=v+a*t 63 current+=s 64 forward_tracks.append(round(s)) 65 66 #反着滑动到准确位置 67 back_tracks=[-3,-3,-2,-2,-2,-2,-2,-1,-1,-1] #总共等于-20 68 69 return {'forward_tracks':forward_tracks,'back_tracks':back_tracks} 70 71 def crack(driver): #破解滑动认证 72 # 1、点击按钮,得到没有缺口的图片 73 button = driver.find_element_by_class_name('geetest_radar_tip') 74 button.click() 75 76 # 2、获取没有缺口的图片 77 image1 = get_image(driver) 78 79 # 3、点击滑动按钮,得到有缺口的图片 80 button = driver.find_element_by_class_name('geetest_slider_button') 81 button.click() 82 83 # 4、获取有缺口的图片 84 image2 = get_image(driver) 85 86 # 5、对比两种图片的像素点,找出位移 87 distance = get_distance(image1, image2) 88 89 # 6、模拟人的行为习惯,根据总位移得到行为轨迹 90 tracks = get_tracks(distance) 91 print(tracks) 92 93 # 7、按照行动轨迹先正向滑动,后反滑动 94 button = driver.find_element_by_class_name('geetest_slider_button') 95 ActionChains(driver).click_and_hold(button).perform() 96 97 # 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 98 for track in tracks['forward_tracks']: 99 ActionChains(driver).move_by_offset(xoffset=track, yoffset=0).perform() 100 101 # 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 102 time.sleep(0.5) 103 for back_track in tracks['back_tracks']: 104 ActionChains(driver).move_by_offset(xoffset=back_track, yoffset=0).perform() 105 106 # 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 107 ActionChains(driver).move_by_offset(xoffset=-3, yoffset=0).perform() 108 ActionChains(driver).move_by_offset(xoffset=3, yoffset=0).perform() 109 110 # 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 111 time.sleep(0.5) 112 ActionChains(driver).release().perform() 113 114 def login_cnblogs(username,password): 115 driver = webdriver.Chrome() 116 try: 117 # 1、输入账号密码回车 118 driver.implicitly_wait(3) 119 driver.get('https://passport.cnblogs.com/user/signin') 120 121 input_username = driver.find_element_by_id('input1') 122 input_pwd = driver.find_element_by_id('input2') 123 signin = driver.find_element_by_id('signin') 124 125 input_username.send_keys(username) 126 input_pwd.send_keys(password) 127 signin.click() 128 129 # 2、破解滑动认证 130 crack(driver) 131 132 time.sleep(10) # 睡时间长一点,确定登录成功 133 finally: 134 driver.close() 135 136 if __name__ == '__main__': 137 login_cnblogs(username='linhaifeng',password='xxxx') 138 139 修订版
用类封装的版本
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
![](https://images.cnblogs.com/OutliningIndicators/ExpandedBlockStart.gif)
1 import time 2 import random 3 4 from selenium.webdriver import ActionChains 5 from selenium.webdriver.common.by import By 6 from PIL import Image 7 8 9 # def simulate_reaction(func): 10 # """模拟人类的反应时间""" 11 # from functools import wraps 12 # 13 # @wraps 14 # def inner(self, *args, **kwargs): 15 # time.sleep(random.uniform(0.2, 1)) 16 # ret = func(self, *args, **kwargs) 17 # return ret 18 # return inner 19 20 21 class SVCR: 22 """识别滑动验证码 极验验证""" 23 24 def __init__(self, driver): 25 self.driver = driver 26 self.get_full_img = True 27 28 # @simulate_reaction 29 def run(self): 30 """执行识别流程""" 31 # 1. 点击按钮开始验证 32 self.click_start_btn() 33 34 # 2. 根据验证类型验证 35 return self.judge_and_auth() 36 37 def judge_and_auth(self): 38 """判断验证类型并执行相应的验证方法""" 39 if True: 40 return self.auth_slide() 41 else: 42 pass 43 44 def auth_slide(self): 45 46 def get_distance(img1, img2): 47 """计算滑动距离""" 48 threshold = 60 49 # 忽略可动滑块部分 50 start_x = 57 51 52 for i in range(start_x, img1.size[0]): 53 for j in range(img1.size[1]): 54 rgb1 = img1.load()[i, j] 55 rgb2 = img2.load()[i, j] 56 res1 = abs(rgb1[0] - rgb2[0]) 57 res2 = abs(rgb1[1] - rgb2[1]) 58 res3 = abs(rgb1[2] - rgb2[2]) 59 if not (res1 < threshold and res2 < threshold and res3 < threshold): 60 return i - 7 # 经过测试,误差为大概为7 61 62 def get_tracks(distance): 63 """ 64 制造滑动轨迹 65 66 策略:匀加速再匀减速,超过一些,再回调,左右小幅度震荡 67 """ 68 69 v = 0 70 current = 0 71 t = 0.2 72 tracks = [] 73 74 # 正向滑动 75 while current < distance+10: 76 if current < distance*2/3: 77 a = 2 78 else: 79 a = -3 80 s = v*t + 0.5*a*(t**2) 81 current += s 82 tracks.append(round(s)) 83 v = v + a*t 84 85 # 往回滑动 86 current = 0 87 while current < 13: 88 if current < distance*2/3: 89 a = 2 90 else: 91 a = -3 92 s = v*t + 0.5*a*(t**2) 93 current += s 94 tracks.append(-round(s)) 95 v = v + a*t 96 97 # 最后修正 98 tracks.extend([2, 2, -3, 2]) 99 100 return tracks 101 102 # 1. 截取完整图片 103 if self.get_full_img: 104 time.sleep(2) # 等待图片加载完毕 105 img_before = self.get_img() 106 else: 107 img_before = self._img_before 108 109 # 2. 点击出现缺口图片 110 slider_btn = self.driver.find_element_by_class_name("geetest_slider_button") 111 slider_btn.click() 112 113 # 3. 截取缺口图片 114 time.sleep(2) # 等待图片加载完毕 115 img_after = self.get_img() 116 117 # 4. 生成移动轨迹 118 tracks = get_tracks(get_distance(img_before, img_after)) 119 120 # 5. 模拟滑动 121 slider_btn = self.driver.find_element_by_class_name("geetest_slider_button") 122 ActionChains(self.driver).click_and_hold(slider_btn).perform() 123 for track in tracks: 124 ActionChains(self.driver).move_by_offset(xoffset=track, yoffset=0).perform() 125 126 # 6. 释放鼠标 127 time.sleep(0.5) # 0.5秒后释放鼠标 128 ActionChains(self.driver).release().perform() 129 130 # 7. 验证是否成功 131 132 time.sleep(2) 133 div_tag = self.driver.find_element_by_class_name("geetest_fullpage_click") 134 if "display: block" in div_tag.get_attribute("style"): 135 '''判断模块对话框是否存在,如果存在就说明没有验证成功,"display: block",重新去验证''' 136 self.get_full_img = False 137 setattr(self, "_img_before", img_before) 138 return self.auth_slide() 139 else: 140 #如果验证成功"display: none" 141 time.sleep(1000) 142 return True 143 144 # @simulate_reaction 145 def click_start_btn(self, search_style="CLASS_NAME", search_content="geetest_radar_tip"): 146 """找到开始按钮并点击""" 147 btn = getattr(self.driver, "find_element")(getattr(By, search_style), search_content) 148 btn.click() 149 150 def get_img(self): 151 """截取图片""" 152 div_tag = self.driver.find_element_by_class_name("geetest_slicebg") 153 154 # 计算截取图片大小 155 img_pt = div_tag.location # {'x': 296, 'y': 15} 156 img_size = div_tag.size # {'height': 159, 'width': 258} 157 img_box = (img_pt["x"], img_pt["y"], img_pt["x"] + img_size["width"], img_pt["y"] + img_size["height"]) 158 159 # 保存当前浏览页面 160 self.driver.save_screenshot("snap.png") 161 162 # 截取目标图片 163 img = Image.open("snap.png") 164 return img.crop(img_box)
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
![](https://images.cnblogs.com/OutliningIndicators/ExpandedBlockStart.gif)
1 from selenium import webdriver 2 3 from svcr import SVCR 4 5 6 def auth(): 7 driver = webdriver.Chrome() 8 # browser.get(url) 9 driver.get("https://passport.cnblogs.com/user/signin") #请求页面 10 driver.implicitly_wait(3) 11 # 第一步:输入账号、密码,然后点击登陆 12 input_name = driver.find_element_by_id('input1') #找到输入用户名的框 13 input_pwd = driver.find_element_by_id('input2') #找到输入密码的框 14 input_button = driver.find_element_by_id('signin') #找到按钮 15 input_name.send_keys("name")#博客园的账号 16 input_pwd.send_keys("pwd")#博客园的密码 17 input_button.click() #进行点击 18 return driver 19 20 def main(): 21 driver=auth() #进行验证, 22 _auth = SVCR(driver) 23 _auth.run() 24 25 if __name__ == '__main__': 26 main()