About
Irin Akther
I research artificial intelligence for cybersecurity — specifically, how we defend the networks behind hospitals, power grids, and water systems.
Background
My work sits at the meeting point of machine learning, network security, and the people who actually have to act on an alert. The technical question is how to make an intrusion-detection model explain its own decisions without giving up capability. The human question — the one I care about most — is who those explanations are written for.
Most of the frontline defending critical infrastructure is not an elite security team. It is thousands of under-resourced operators who receive the same machine-generated alert as a veteran analyst and are expected to act on it. Building AI that meets them where they are is the throughline of everything below.
Focus areas
Explainable AI (XAI)
Feature-level evidence behind every alert, evaluated on public benchmarks including CICIDS2017.
Network intrusion detection
Detection models for operational-technology and critical-infrastructure networks.
Human-centered security
Persona-adaptive, on-device explanations — a patent-pending (provisional) approach.
How I work
- Defensive only. This is protective research for critical infrastructure. It does not produce offensive capability.
- Claims stay inside the evidence. Benchmark performance is reported as benchmark performance — never as proof of real-world generalization.
- Limitations are part of the result. Performance-versus-transparency tension, single-dataset scope, and adaptive adversaries get named, not buried.
- Privacy by construction. Where explanations can be generated on-device, sensitive operational data never has to leave the network.
Topics I speak and write about
- Why “threat detected” is not enough — alert fatigue as a trust problem
- Black box to glass box: explainability in intrusion detection
- Explanations designed for the reader's technical literacy
- Honest trade-offs in trustworthy AI for security
- Explainability as critical-infrastructure policy, not a feature request
Contact
Open to research collaboration, peer review, speaking on explainable AI for critical-infrastructure defense, and conversations with operators and security teams about what would actually help on the ground.
- linkedin.com
- Based in
- United States