DIREC project

Training Robust Network Intrusion Detection with Differentiable Logics

An Intelligent Defense Against Hackers

Project impact

Normally, data flows safely through the network. But a cyberattack can change everything—often without being detected. In many cases, it can look like perfectly normal behavior, only differing slightly in timing or sequence.

A traditional NIDS (Network Intrusion Detection System) based on synthetic data often overlooks such subtle attacks. However, by adding differentiable logics—trained on real-world data and grounded in security rules—the system can alert you in time, even when attacks are deliberately camouflaged.

In this research project, scientists from ITU, DTU, AAU, and the cybersecurity company LogPoint will combine the flexibility of machine learning with the rigor of logic. By developing differentiable logics, the monitoring system gains a kind of digital intuition that can “think” outside the box and uncover manipulations from hostile systems.

PROJECT DATA

Project name

Training Robust Network Intrusion Detection with Differentiable Logics

Project period

2026-2027

Funding

DKK 2.000.000

Scientific mission

The project builds on the NFC project “Robust AI Algorithms for Network Intrusion Detection” from 2025, which developed the first prototype of a NIDS using differentiable logics.

The next step is to test the technology in collaboration with LogPoint, which provides testing facilities and access to customers. This is an important complement to ongoing research, offering the opportunity to evaluate NIDS with differentiable logics in an environment that researchers normally do not have access to.

To develop the prototype into a solution ready for practical use, the researchers will identify and integrate a comprehensive set of new features into the product. Researchers from ITU focus on formalizing the differentiable logics, while DTU and AAU are responsible for further developing the detection systems and integrating artificial intelligence.

The team will collect training data for relevant attacks, and LogPoint will test the final detection mechanism in commercial operation. The ambition is to close the current gap between theory and implementation, so that the technology can scale beyond the laboratory and strengthen cybersecurity across society.

Project Participants

	Alessandro Bruni – Associate Professor – IT University Copenhagen
	Nicola Dragoni – Professor – Technical University of Denmark
	Giorgio Bacci – Associate Professor – Aalborg University
	Gust Grinbergs – Research Assistant – IT University Copenhagen
	Amine Laghaout – R&D Project Manager – Logpoint