InnoCaptcha

InnoCaptcha is a professional, pluggable Python CAPTCHA library designed for modern web applications. It supports multiple challenge types, from traditional text and math to advanced audio, image-based grid challenges (using YOLOv11), and voice recognition challenges.

---

🌟 Key Features

• 📝 Text CAPTCHA: Highly configurable image-based text challenges with anti-aliasing and distortion.
• 🔢 Math CAPTCHA: Secure arithmetic problems (no eval()) with optional image rendering.
• 🎧 Audio CAPTCHA: Spoken character sequences with noise injection and variable speed.
• 🗣️ Voice CAPTCHA (STT): Speech-to-text challenges where users must speak a random phrase.
• 🖼️ Image CAPTCHA (YOLOv11): Advanced 3×3 grid challenges using Object Detection.
• 🔐 Security First:
  • Token-based challenge identification.
  • IP and Session binding for verification safety.
  • Automatic expiration (5 minutes) and attempt limits (5 attempts).
  • Synchronous cleanup for expired challenges to prevent memory leaks.
  • Secure data encryption via INNOCAPTCHA_KEY.
• 🗄️ Storage: Centralized SQLite database management.

---

🚀 Installation

pip install InnoCaptcha

🛠️ Quick Start

Important: For production, please set the INNOCAPTCHA_KEY environment variable to a secure Fernet key to encrypt answers in the database.

export INNOCAPTCHA_KEY="your_secure_fernet_key_here"

1. Text CAPTCHA

Generates a distorted image containing a random string.

from InnoCaptcha.text import TextCaptcha

captcha = TextCaptcha(width=300, height=80)
# Pass IP and session ID for context binding
captcha_id = captcha.create("abcd", ip="127.0.0.1", session_id="abc123xyz")
captcha.save("captcha.png")

print(captcha.verify("abcd", ip="127.0.0.1", session_id="abc123xyz")) # Returns True

2. Math CAPTCHA

Generates arithmetic challenges. Can be output as plain text or a rendered image.

from InnoCaptcha.math import MathCaptcha

# Image-based Math Challenge
math_cap = MathCaptcha(output="image")
math_cap.create(ip="127.0.0.1", session_id="abc123xyz")
math_cap.get_question().show() # Returns a PIL Image

# Answer depends on the generated math question
print(math_cap.verify("12", ip="127.0.0.1", session_id="abc123xyz"))

3. Audio CAPTCHA

Generates a WAV file where a voice reads out characters.

from InnoCaptcha.audio import AudioCaptcha

audio = AudioCaptcha()
audio.create("x123", ip="127.0.0.1")
audio.save("output.wav")

print(audio.verify("x123", ip="127.0.0.1"))

4. Voice CAPTCHA (New!)

A speech-to-text challenge. The user is given a phrase and must submit a recording of them speaking it.

from InnoCaptcha.voice import VoiceCaptcha

vc = VoiceCaptcha(language='en-US')
captcha_id = vc.create(ip="127.0.0.1") 
print(f"Please read: {vc.phrase}")

# ... User records audio and sends bytes ...
audio_bytes = open("user_speech.wav", "rb").read()
is_correct = vc.verify(audio_bytes, ip="127.0.0.1")

5. Image CAPTCHA (YOLOv11)

Uses YOLOv11 to detect objects in an image and asks the user to select the grid cells (1-9).

from InnoCaptcha.image import ImageCaptcha

img_cap = ImageCaptcha()
img_cap.create(ip="127.0.0.1")
img_cap.save("grid_image.png")
print(f"Target object to detect: {img_cap.image_class}")

# User inputs cell numbers, e.g., "1,2,5"
print(img_cap.verify("1,2,5", ip="127.0.0.1"))

---

💎 API Reference

Component	Description
TextCaptcha	Classic image-text challenges. Supports custom colors/scaling.
MathCaptcha	Arithmetic challenges (+, -, *). Supports text or image output.
AudioCaptcha	Generates character-based audio files for auditory verification.
VoiceCaptcha	Speech-to-text verification using speech_recognition.
ImageCaptcha	AI-powered grid identification using YOLOv11.

---

🆙 Latest Updates (v2.3.x)

• New Security & Architecture Audit Fixes:
  • Addressed memory leaks by removing background threading loops per instance.
  • Patched an SQL injection via ALLOWED_TABLES whitelisting.
  • Eliminated timing attacks across all context verifications using secrets.compare_digest.
  • Moved DB Encryption to require an INNOCAPTCHA_KEY variable, keeping the encryption key out of the SQLite DB itself.
  • Implemented atomic updates for attempt counters to eliminate race conditions.
  • Optimized ImageCaptcha to use a global cached singleton for its YOLOv11 model, vastly speeding up subsequent initializations.

---

📜 Requirements

• Python 3.9+
• Dependencies: Pillow, numpy, scipy, ultralytics, opencv-python, pydub, SpeechRecognition, etc.

---

📄 License

MIT © InnoSoft Company