Webinar – Can Machine Learning Keep a Secret? Privacy Risks and Protection TechniquesWebinar –

Machine learning is increasingly used in areas that rely on sensitive and personal data, such as healthcare, finance, and personalised digital services. While modern models can achieve very high accuracy, they also raise important privacy concerns. Even when training data are not directly shared, models may still reveal information about individuals through their predictions or internal representations. This lecture offers an introduction to privacy issues in machine learning and explains why they matter in practice.

The lecture starts by exploring the main privacy risks in machine learning, focusing on how models can reveal information about their training data. Examples of common attacks, including membership inference, attribute inference, and model inversion, help illustrate why standard data protection and anonymisation approaches are often inadequate in the context of machine learning.

Building on this motivation, the lecture presents the key approaches for protecting privacy in modern machine learning. These include differential privacy for controlling information leakage, as well as distributed learning methods like federated learning. It also introduces cryptographic and system-level techniques, such as secure aggregation and encrypted inference, which help limit data exposure and reduce trust in centralised systems.

By the end of the lecture, participants will have a clear overview of where privacy risks in machine learning come from and how different privacy-preserving techniques can be combined to build more trustworthy AI systems.

Sasho Gramatikov

Bio:

Dr. Sasho Gramatikov is a Full Professor at the Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University in Skopje. He teaches undergraduate and master courses in Web Programming, Web Security, Privacy in Machine Learning, and other core computer science subjects. He holds a Ph.D.  from Universidad Politécnica de Madrid (Cum Laude), along with an M.Sc. and B.Sc. in Computer Engineering from Ss. Cyril and Methodius University.

His research centres on machine learning, applied AI, linked data, and web technologies, with recent work focusing on AI applications in food systems, medicine, and practical, industry-oriented use cases. He is involved in several international projects, including ChatMED, EuroCC2, FoodMarketMap, and MultiplEYE, which address topics such as generative AI in healthcare, high-performance computing, personalised nutrition, and multilingual eye-tracking and language processing.