Hours:
16 hours (4 credits)

Room:

Aula Riunioni del Dipartimento di Ingegneria dell’Informazione, Via G. Caruso 16, Pisa – Ground Floor e Aula Riunioni Piano 6 - L. Lazzarino

To register to the course, click here

Short Abstract:

This course introduces generative modelling approaches for anomaly detection in cybersecurity systems. Modern intrusion detection increasingly relies on modelling the probability distribution of normal system behaviour and identifying deviations from this distribution. Generative models (including probabilistic models, autoencoders, variational approaches, and adversarial learning) provide a principled framework for learning such representations from complex data.

The course will cover the theoretical foundations of generative modelling, the role of probability distributions in anomaly detection, and practical approaches to applying generative models to cybersecurity data. Students will gain both conceptual understanding and practical experience implementing generative models for intrusion detection.

Course Contents in brief:

1. Foundations of anomaly detection and intrusion detection systems
2. Probabilistic modelling of normal system behaviour
3. Classical generative models for anomaly detection
4. Autoencoders and deep generative architectures
5. Variational Autoencoders and latent variable models
6. Generative Adversarial Networks
7. Evaluation of generative intrusion detection systems

Schedule:

1. Day 1 (Monday, 13th April, 9.30 – 12.30): Foundations of Intrusion Detection and Anomaly Detection (3 hours) - Via Caruso
2. Day 2 (Tuesday, 14th April, 9.30 – 12.30): Generative Modelling Foundations (3 hours) - Via Caruso
3. Day 3 (Wednesday, 15th April, 9.30 – 12.30): Deep Generative Models for Anomaly Detection (3 hours) - L. Lazzarino
4. Day 4 (Thursday, 16th April, 12.30 – 15.30): Advanced Generative Approaches (3 hours) - L. Lazzarino
5. Day 5 (Friday, 17th, 9.00 – 13.00): Practical Applications (4 hours) Via Caruso