Hybrid GCN + Multi-Head Attention + DQNEnsemble ML

Intrusion Detection
Inference Platform

Run secure predictions on UNSW (Attack vs Normal) and CIC (Normal, DoS, PortScan) with deep learning and production-grade ensemble ML models.

Start InferenceAPI Documentation

Datasets

UNSW-NB15Binary Classification — Attack / Normal
CIC-IDS 20173-Class — Normal / DoS / PortScan

Deep Learning

Hybrid Graph Convolutional Network with Multi-Head Attention and Deep Q-Network. Checkpointed models per dataset.

GCN → Attention → DQN Policy

Machine Learning

Tree-BasedRandom Forest, Extra Trees
Gradient BoostingXGBoost, LightGBM, CatBoost

Features

Real-time JSON OutputCSV & JSON UploadsSingle Sample InputBatch ProcessingConfidence Scores