Teaching Cybersecurity in the Age of AI

How our recent paper on AI-based intrusion detection is reshaping the way we teach networks and security to MCA students.

For two decades our networks course has followed the same rhythm: TCP/IP, firewalls, IDS, a project. This year we tore it up.

Our recent paper on adaptive AI-based intrusion detection made one thing plain: the interesting failure modes in modern networks are behavioural, not signature-based. Students who graduate today should be able to reason about anomaly detection, not just configure a rule.

The new syllabus interleaves classical networking with three practical AI modules — feature engineering on flow data, model training on public benchmarks, and, crucially, a red-team exercise in which teams try to evade one another's detectors.

Early signal is good: the first cohort's capstone projects are noticeably more ambitious, and two are being written up for undergraduate research venues.