Миша Ефремов • Dev.to

Миша Ефремов (individual developer)

-

SHARD system integrating XGBoost, Seq2Seq Transformer, RL DQN Agent, VAE Anomaly Detector, GNNs

Small businesses face frequent cyberattacks such as SQL injections, brute force attempts, DDoS floods, and ransomware but cannot afford costly enterprise SIEM solutions. There is a need for an automated, effective defense system that operates without a dedicated security team.

Detect cyberattacks on servers, classify attack types, generate and apply real-time defense rules, block or throttle attackers, detect zero-day attacks, predict attacker next moves, and notify administrators automatically.

13 honeypots for attack detection (SSH, MySQL, Redis, MongoDB, FTP, etc.), XGBoost for attack classification, Seq2Seq Transformer for rule generation, Reinforcement Learning DQN agent for defense decisions, Variational Autoencoder for anomaly detection, Graph Neural Networks for threat mapping and prediction, Multi-Modal Fusion for threat scoring, PyTorch, Docker, Swagger, pytest, Telegram API for notifications.