MAFIA - Middleware AI Firewall Application¶

MAFIA is a Python middleware designed to detect and prevent brute force attacks and block malicious payloads (such as XSS, SQL Injection, etc.). This middleware can be easily integrated into FastAPI or Starlette-based applications.

🔧 Instalation¶

pip install git+https://github.com/akarkode/mafia.git

🚀 Usage¶

Add the middleware to your FastAPI application:

from mafia.middleware import BruteForceMiddleware
from fastapi import FastAPI

app = FastAPI()
app.add_middleware(BruteForceMiddleware, redis_url="redis://localhost:6379")

This middleware will automatically: - Detect and block suspicious requests from the same IP. - Classify login attempts using an AI-based model. - Block any requests containing potentially malicious payloads.

🛡️ Security Features¶

1. Brute Force Detection¶

The BruteForceDetector AI model analyzes login attempt patterns based on frequency and interval features.

2. Malicious Payload Filtering¶

The SecurityFilter scans and blocks: - XSS attacks: <script>, onerror=, javascript: - SQL Injection: ' OR 1=1, UNION SELECT, --, etc.

3. Rate Limiting¶

The RedisClient module limits request frequency based on the user's IP address.

🧠 Internal Architecture¶

[Client] → [BruteForceMiddleware]
   ├─ [RedisClient] - rate limiting
   ├─ [BruteForceDetector] - brute force classification
   └─ [SecurityFilter] - XSS/SQLi filtering

Key Classes and Functions¶

BruteForceMiddleware
dispatch(): Core middleware process
BruteForceDetector
predict(features: list) -> bool
train(), save_model()
SecurityFilter
is_malicious(payload: str) -> bool

📈 Future Improvements¶

Real-time log visualization dashboard.
Integration with ElasticSearch for advanced logging.
LSTM-based AI model for more accurate login pattern recognition.
Periodic automatic training of the detection model.

📄 License and Contribution¶

Please refer to the LICENSE file for license details. Pull requests, issues, and contributions are very welcome.

This documentation is built using MkDocs Material as a single page for simplicity and fast access.