Detecting an attack is only half the battle; security teams also need to see quickly where it is happening. Neural-Heatmap turns AI intrusion detection into a live, colour-coded map of risk across an organisation’s divisions.
The problem
Many AI intrusion detection studies focus on accuracy but give little attention to real-time visualisation that helps people respond. In large, multi-division organisations, finding the compromised area is slow.
What the research did
We combined a BiLSTM intrusion classifier with a heatmap-based risk monitoring interface. The system processes network traffic, estimates anomaly probability and maps each division’s risk level onto a real-time visual dashboard.
What we found
On the CIC-IDS2017 dataset, Neural-Heatmap reached 98.86% average accuracy, 98.82% precision, 98.76% recall and a 98.79% F1-score under 10-fold cross-validation. It cut compromised-zone tracking time by 89.82% compared with log-based inspection.
Who benefits
- Security operations teams in large enterprises
- IT managers who need a clear overview
- Organisations with many sites or departments
Limits and what comes next
Evaluated on a benchmark dataset; real enterprise traffic and layouts vary, so pilot deployments are the next step.
The team
This research was led by N. V. D. S. S. V. Prasad Raju (BITS Pilani) with me, Moutaz Alazab and Olivia-Roxana Alecsoiu. It was presented at ICAISET 2026 in Cairo.
Frequently asked questions
What is Neural-Heatmap?
A system that combines AI intrusion detection with a live risk map of an organisation.
How accurate is it?
98.86% average accuracy on CIC-IDS2017.
How does it help responders?
It cut the time to locate compromised zones by 89.82%.