🏗️ Methodology

Data Pipeline

1. Logs are collected from a relational database.
2. Features include:
   • Timestamps
   • IP addresses
   • Request types
   • Geolocation data
3. Data is cleaned to remove duplicates and irrelevant logs.

Neural Network Architecture

Model Selection:

• LSTMs (Long Short-Term Memory) for sequential pattern detection.
• CNNs (Convolutional Neural Networks) for pattern-based anomaly detection.

Training Process:

• Supervised Learning: Labeled logs for attack/non-attack detection.
• Unsupervised Learning: Autoencoders for detecting unknown attack patterns.

Deployment Plan

• Model integrates with a dashboard.
• Real-time anomaly alerts.
• API for external monitoring tools.

🚀 Stay tuned for more updates!