Avviso Seminario Danilo Francati
Date: Wednesday, May 14
Time: 09.00
Venue: https://meet.google.com/roc-cugq-sxj
Speaker: Danilo Francati
Titolo: Security Through Cryptography: From Computing Predicates to Trustworthy Generative AI.
Abstract: In recent years, cryptographic primitives have undergone a remarkable transformation, evolving from tools for secure communication into powerful mechanisms for solving increasingly complex problems.
In this talk, we will trace the evolution of cryptography through the lens of our growing need for privacy-preserving systems. Starting from foundational concepts such as predicate encryption, we will explore how this framework has expanded to support sophisticated functionalities, including matchmaking encryption and multi-input predicate encryption. These tools enable fine-grained, policy-based access to encrypted data, fundamentally reshaping our approach to secure communications.
Then, we will shift focus to the emerging role of generative AI, which is reshaping our societies and forcing a reevaluation of our security needs. We will examine how current cryptographic techniques respond to these new challenges, highlighting both their potential and their limitations in addressing the risks posed by AI-generated content and models..
In this talk, we will trace the evolution of cryptography through the lens of our growing need for privacy-preserving systems. Starting from foundational concepts such as predicate encryption, we will explore how this framework has expanded to support sophisticated functionalities, including matchmaking encryption and multi-input predicate encryption. These tools enable fine-grained, policy-based access to encrypted data, fundamentally reshaping our approach to secure communications.
Then, we will shift focus to the emerging role of generative AI, which is reshaping our societies and forcing a reevaluation of our security needs. We will examine how current cryptographic techniques respond to these new challenges, highlighting both their potential and their limitations in addressing the risks posed by AI-generated content and models..