gpt4-captcha-bypass/main.py

import argparse
import os
import time
import random
import requests
import re
from dotenv import load_dotenv
from selenium import webdriver
from selenium.webdriver import ActionChains
from selenium.webdriver.common.by import By
from selenium.webdriver.firefox.service import Service as FirefoxService
from selenium.webdriver.support.ui import WebDriverWait
from selenium.webdriver.support import expected_conditions as EC
from webdriver_manager.firefox import GeckoDriverManager
from openai import OpenAI

load_dotenv()
def average_of_array(arr):
    if not arr:
        return 0  # Handle edge case of empty array
    sum_elements = sum(arr)
    average = sum_elements / len(arr)
    return average - 5

def upload_image_to_imgur(image_path):
    url = "https://api.imgur.com/3/image"
    client_id = os.getenv("IMGUR_CLIENT_ID")
    headers = {
        "Authorization": f"Client-ID {client_id}"
    }

    with open(image_path, "rb") as image_file:
        payload = {
            "image": image_file.read(),
            "type": "file"
        }
        response = requests.post(url, headers=headers, files=payload)

    if response.status_code == 200:
        data = response.json()
        return data['data']['link']
    else:
        print("Failed to upload image", response.text)
        return None

def ask_recaptcha_to_chatgpt(image_url):
    client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
    short_prompt = ("You are an object detection assistant. Understand what is asked on the top instruction of the "
                    "image. For Example if instruction says select all squares with motorcycle. There are 16 squares "
                    "give number for each square 0 to 15. The numbers starts from top left to right. Then answer only "
                    "with the square numbers like 1-3-5-6."
                    "Understand the instruction and give me the highest probability squares as answer. Give only the correct square numbers.")
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": short_prompt},
            {"role": "user", "content": image_url}
        ],
        temperature=1,
        max_tokens=256,
        top_p=1,
        frequency_penalty=0,
        presence_penalty=0
    )
    return response.choices[0].message.content

def ask_text_to_chatgpt(image_url):
    client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
    short_prompt = ("Act as a blind person assistant. Read the text from the image and give me only the text answer.")
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {
                "role": "system",
                "content": [
                    {
                        "type": "text",
                        "text": short_prompt
                    }
                ]
            },
            {
                "role": "user",
                "content": [
                    {
                        "type": "image_url",
                        "image_url": {
                            "url": image_url
                        }
                    },
                    {
                        "type": "text",
                        "text": "Understand what is asked on the top instruction of the image. Give me the only correct squares with highest probability"
                    }
                ]
            },
        ],
        temperature=1,
        max_tokens=256,
        top_p=1,
        frequency_penalty=0,
        presence_penalty=0
    )
    return response.choices[0].message.content

def ask_slide_to_chatgpt(image_url):
    client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
    short_prompt = ("There is an slider button in the page and there is an empty gray space on puzzle. You should give "
                    "give me how many pixels should i slide to complete the puzzle. You can simulate the slider for "
                    "the exact fit before giving the answer.The total width is 210 pixels."
                    "Give me only number of pixels in integer 50 up to 210 pixels.")
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": short_prompt},
            {"role": "user", "content": image_url}
        ],
        temperature=1,
        max_tokens=256,
        top_p=1,
        frequency_penalty=0,
        presence_penalty=0
    )
    return response.choices[0].message.content

def puzzle_test(driver):
    driver.get("https://2captcha.com/demo/geetest")
    time.sleep(5)
    button = WebDriverWait(driver, 10).until(
        EC.element_to_be_clickable((By.CLASS_NAME, "geetest_radar_tip"))
    )
    button.click()
    box = WebDriverWait(driver, 10).until(
        EC.presence_of_element_located((By.CLASS_NAME, "geetest_embed"))
    )
    time.sleep(1)
    box.screenshot('geetest_box.png')
    file_url = upload_image_to_imgur('geetest_box.png')
    time.sleep(1)
    all_results = []
    while True:
        slider = WebDriverWait(driver, 10).until(
            EC.presence_of_element_located((By.CLASS_NAME, "geetest_slider_button"))
        )
        time.sleep(2)
        action = ActionChains(driver)
        input_string = ask_slide_to_chatgpt(file_url)
        numbers = re.findall(r'\d+', input_string)
        if numbers:
            result = int(numbers[0])
        else:
            result = None
        if result < 110:
            result = 90
        all_results.append(result)
        action.click_and_hold(slider).perform()
        time.sleep(random.uniform(0.8, 1.2))
        total_offset = average_of_array(all_results)
        num_steps = 5
        step_offset = total_offset / num_steps
        for _ in range(num_steps):
            action.move_by_offset(step_offset, 0).perform()
            time.sleep(random.uniform(0.05, 0.1))
        action.release().perform()

def complicated_text_test(driver):
    driver.get("https://2captcha.com/demo/mtcaptcha")
    time.sleep(5)
    iframe = WebDriverWait(driver, 10).until(
        EC.presence_of_element_located((By.ID, "mtcaptcha-iframe-1"))
    )
    iframe.screenshot('captcha_image.png')
    file_url = upload_image_to_imgur('captcha_image.png')
    print(file_url)
    response = ask_text_to_chatgpt(file_url)

    print(response)
    driver.switch_to.frame(iframe)
    input_field = WebDriverWait(driver, 10).until(
        EC.presence_of_element_located((By.CLASS_NAME, "mtcap-noborder.mtcap-inputtext.mtcap-inputtext-custom"))
    )
    input_field.send_keys(response)
    time.sleep(2)
    driver.switch_to.default_content()
    submit_button = WebDriverWait(driver, 10).until(
        EC.presence_of_element_located(
            (By.CLASS_NAME, "_actionsItem_1f3oo_41._buttonPrimary_d46vc_44._button_d46vc_1._buttonMd_d46vc_34"))
    )
    submit_button.click()

def text_test(driver):
    driver.get("https://2captcha.com/demo/normal")
    time.sleep(5)
    captcha_image = WebDriverWait(driver, 10).until(
        EC.presence_of_element_located((By.CLASS_NAME, "_captchaImage_rrn3u_9"))
    )

    time.sleep(2)
    captcha_image.screenshot('captcha_image.png')
    file_url = upload_image_to_imgur('captcha_image.png')
    print(file_url)
    response = ask_text_to_chatgpt(file_url)

    print(response)

    input_field = WebDriverWait(driver, 10).until(
        EC.presence_of_element_located((By.CLASS_NAME, "_inputInner_ws73z_12"))
    )
    input_field.send_keys(response)
    submit_button = WebDriverWait(driver, 10).until(
        EC.presence_of_element_located(
            (By.CLASS_NAME, "_actionsItem_1f3oo_41._buttonPrimary_d46vc_44._button_d46vc_1._buttonMd_d46vc_34"))
    )
    submit_button.click()
    time.sleep(5)
    driver.quit()

def recaptcha_test(driver):
    driver.get("https://patrickhlauke.github.io/recaptcha/")
    def handle_recaptcha():
        number_of_challenges = 0
        recaptcha_frame = WebDriverWait(driver, 10).until(
            EC.presence_of_element_located((By.XPATH, "//iframe[@title='reCAPTCHA']"))
        )
        recaptcha_frame.screenshot('recaptcha_checkbox.png')
        driver.switch_to.frame(recaptcha_frame)
        checkbox = WebDriverWait(driver, 10).until(
            EC.element_to_be_clickable((By.CLASS_NAME, "recaptcha-checkbox-border"))
        )
        checkbox.click()
        driver.switch_to.default_content()
        time.sleep(2)
        array = []
        while True:
            try:
                challenge_iframe = WebDriverWait(driver, 10).until(
                    EC.presence_of_element_located(
                        (By.XPATH, "//iframe[contains(@title, 'recaptcha challenge expires in two minutes')]"))
                )
                number_of_challenges += 1
                filename = 'recaptcha_challenge_' + str(number_of_challenges) + '.png'
                challenge_iframe.screenshot(filename)
                image_url = upload_image_to_imgur(filename)
                if image_url:
                    answer = ask_recaptcha_to_chatgpt(image_url)
                    if "," in answer:
                        array = answer.split(", ")
                    elif "-" in answer:
                        array = answer.split("-")
                else:
                    exit(-1)
                driver.switch_to.frame(challenge_iframe)
                table = WebDriverWait(driver, 10).until(
                    EC.presence_of_element_located((By.CLASS_NAME, "rc-imageselect-table-33"))
                )
                time.sleep(1)
                all_images = table.find_elements(By.CLASS_NAME, "rc-image-tile-33")
                for each_element in array:
                    all_images[int(each_element)].click()
                submit_button = WebDriverWait(driver, 10).until(
                    EC.presence_of_element_located((By.ID, "recaptcha-verify-button"))
                )
                submit_button.click()
                driver.switch_to.default_content()
                time.sleep(5)
                if len(driver.find_elements(By.CLASS_NAME, "recaptcha-checkbox-checkmark")) != 0:
                    break
            except Exception as ex:
                print(ex)
    handle_recaptcha()
    time.sleep(5)
    driver.quit()

def main():
    parser = argparse.ArgumentParser(description="Test various captcha types.")
    parser.add_argument('captcha_type', choices=['puzzle', 'text', 'complicated_text', 'recaptcha'],
                        help="Specify the type of captcha to test")
    args = parser.parse_args()

    service = FirefoxService(GeckoDriverManager().install())
    driver = webdriver.Firefox(service=service)
    try:
        if args.captcha_type == 'puzzle':
            puzzle_test(driver)
        elif args.captcha_type == 'text':
            text_test(driver)
        elif args.captcha_type == 'complicated_text':
            complicated_text_test(driver)
        elif args.captcha_type == 'recaptcha':
            recaptcha_test(driver)
    finally:
        driver.quit()

if __name__ == "__main__":
    main()