Automata learning has become a popular technique to facilitate analysis software and hardware behavior. It is particularly suited for behaviors that can be described as finite state machines, as for example network protocols, but can be used in a surprisingly wide range of applications. In this talk, I want to explore the limits of using automata learning in the context of network verification. I will give an overview of recent work on the use of active automata learning to analyse QUIC, a network protocol that is bound to eventually replace TCP. I will then discuss current work on active learning for weighted and probabilistic automata, needed in order to analyse more expressive network behaviors, e.g. congestion and fault tolerance.
25/11/2024
The talk will take place on Monday November 25, at 12:00pm CET in Room 101 (Palazzina D, Viale Regina Elena 295). Mark it down in your agenda.