Advanced Topics in Ai and Cybersecurity

The cluster aims to provide students with integrated and in-depth knowledge of selected topics in artificial intelligence, with particular reference to deep learning, and cybersecurity, building on the competencies introduced in previous clusters. It combines model development and experimental evaluation with the analysis of security threats in controlled and authorised environments, strengthening a methodological approach to experimentation.
The cluster introduces and consolidates the foundations of deep learning, with reference to major architectures and training strategies, also considering robustness aspects and basic security requirements in model design and evaluation. In parallel, it addresses offensive security principles and methodologies exclusively for educational purposes, focusing on vulnerability identification and attack-behaviour analysis in dedicated environments, in compliance with authorisation constraints and responsible practices.
The cluster is organised into the modules Deep Learning, Offensive Security, and AI Laboratory, which are designed in a coherent and coordinated manner. The modules jointly contribute to the intended learning outcomes by integrating data/model experimentation, attack-surface analysis, and mitigation reasoning, with attention to activity traceability and the production of essential technical documentation.

Knowledge and understanding
At the end of the cluster, the student acquires knowledge required to understand deep learning methods and offensive-security workflows in controlled contexts. In particular, the student is able to:
· describe key deep learning concepts and the role of major architectures and training techniques in modern applications;
· outline the phases and objectives of an offensive-security process for educational purposes (reconnaissance, vulnerability analysis, validation), and the role of controlled environments and authorisations.
Applying knowledge and understanding
At the end of the cluster, the student is able to:
· set up datasets, objectives, and evaluation procedures to train and assess deep learning models using standard tools and libraries, in laboratory settings;
· conduct guided offensive-security activities in controlled and authorised environments, collecting evidence and relating observations to potential mitigations;
· organise reproducible experimental workflows (code, data, configurations) and produce structured technical artefacts (e.g., notebooks and short reports).
Making judgements
The student develops the ability to:
· interpret model performance and recognise common issues (e.g., overfitting/underfitting), proposing justified corrective actions;
· critically assess evidence gathered during guided offensive-security activities, identifying limitations, residual risks, and trade-offs of mitigation options.
Communication skills
At the end of the cluster, the student is able to:
· document experiments and results (AI and security) in a clear and technically correct manner, including assumptions, limitations, and operational implications.
Learning skills
The student acquires the ability to:
· consult and synthesise specialised resources (research papers, technical documentation) to reproduce, adapt, or extend deep learning methods and security techniques in controlled contexts;
· autonomously plan experiments and validations, ensuring traceability and reproducibility (versioning of code, data, configurations) and critically evaluating results;
· update analysis and testing tools/methods responsibly, operating only in authorised contexts and in compliance with ethical and legal principles.