Artificial intelligence: secure and robust design

Home page
Artificial intelligence: secure and robust design

This module aims at exploring the various stages involved in designing an AI-based approach across a range of application areas, including cybersecurity, whilst highlighting the security threats involved. Guidelines for safe design and tools for safety assessment will be provided. 

Would you like to attend the course live with the tutor? Download the course timetable and get in touch

Topics

An overview of the various approaches based on Artificial Intelligence. Machine Learning, Deep Learning, generative Models. Supervised learning and unsupervised learning. Critical aspects in defining the data model, managing the training set, selecting the learning algorithm and estimating parameters, as well as challenges during the operational phase. Threat models definition. Robust design techniques. Tools for assessing the security and robustness of the developed system.

Target

  • Software development companies or third-party solution integrators.
  • Designers and software developers.

Course structure

Module 1 – BASICS OF AI, MACHINE LEARNING AND DEEP LEARNING
8 hours
  • An introduction to the fundamental concepts
  • Generative models and applications
  • Potential cybersecurity risks
Module 2 - DATASET, DATA MODELS
8 hours
  • Data pre-processing: cleaning, standardisation and handling of missing values
  • Dataset features
  • Data models and training sets
Module 3 – METHODS FOR USING PARAMETERS
8 hours
  • Techniques for parameter selection
  • Cross-validation and grid search
  • Methods for evaluating model performance
  • Model evaluation
  • Performance metrics (precision, recall, F1-score, ROC curve)
  • Error analysis and iterative improvement
Module 4 – Types of Errors and Threat Modelling
8 hours
  • Threat models for AI systems
  • Vulnerability analysis based on breaches of confidentiality, integrity and availability (CIA)
  • Impact on data integrity
  • Manipulation of input data (e.g., adversarial attacks);
  • Robustness against noisy or manipulated data
Module 5 – PROBLEM MITIGATION TECHNIQUES
8 hours
  • Strategies for mitigating threats (Regularisation and dropout for robust models)
  • Best practices for secure design
  • Principles of security by design
  • Creating a secure AI model
  • Application of model hardening techniques to improve robustness
  • Validation and testing of the designed model
  • Continuous monitoring and auditing of models
WORKSHOP
16 hours
  • Practical exercises in designing machine learning systems, vulnerabilities, and robust and secure design across a range of different application areas.
  • Review what we have covered in theory and apply it to a case study. Data sets, design, risks and resolution.

Artificial intelligence: secure and robust design

Index

Request more information

Request more information

Fill in all sections of the form and click CONFIRM.

You may also be interested in…

Safety considerations and relevant regulations in the automotive industry

Security aspects in embedded devices

Security and privacy on mobile devices
  • Dal 31 ottobre al 5 dicembre 2025
  • Instructor: Prof. Alessio Merlo

Software security: preventing vulnerabilities through secure programming
  • dal 20 ottobre al 24 novembre 2025