FORECAST YOURTRAFFIC INTO PREDICTED
ATTACK STATES.

TEAM COD-I

↓ SCROLL DOWN

IIIT BHAGALPUR

PROBLEM STATEMENT

Traditional machine learning classifiers applied to network traffic treat each flow in isolation. We need AI systems capable of learning network behaviour, anticipating attacker progression and supporting proactive cyber defence using the emerging concept of World Models.

NTRO CHALLENGE

PROCESS

HOW IT WORKS

Here's how we build solutions that actually work:

LEARN MORE

INGEST

Ingest Network Data

We are ingesting Network Data and packets continuously to capture every detail and temporal behavior of the traffic flow.

MODEL

GNN and TFT Models

The ingested models go to Graph Neural Networks (GNN) and Temporal Fusion Transformers (TFT) to learn complex transition dynamics.

ACT

Dashboard and Automation

Output to a score dashboard for interpretable decision support, with possible automated defensive measures triggered dynamically.

OUR SERVICES

THREE OPTIONS. ONE GOAL.

Network Ingestion

Real-Time

A streamlined data pipeline that ingests flow-level and packet-level features. Captures aggregate behavior and sequencing patterns designed by attackers to evade thresholds.

Flow-Level Vectors,Packet-Level Timing,Feature Engineering,Telemetry Parsing

World Model

Continuous

Learn state-transition dynamics using advanced AI. Given the current observed network state, the model calculates the probability distribution over future states.

Graph Neural Networks,Temporal Fusion,Transition Dynamics,Forward Simulation

Predictive Defence

Proactive

Forecast future network states and estimate the probability of attacker progression. Maps predicted behavior to MITRE ATT&CK stages and provides actionable explainability.

Threat Forecasting,MITRE Mapping,Attention Mechanisms,Decision Support

FORECAST YOURTRAFFIC INTO PREDICTEDATTACK STATES.