Amaury BEAUDET defended his PhD on June 7th, 2024.
Place : amphithéâtre Clémence Royer, département Génie Mécanique, INSA Lyon, 27 avenue Jean Capelle Ouest, 69621 Villeurbanne"
Jury :
Rapporteurs :
M. LEFEBVRE, Dimitri, Professeur des universités, Université de Normandie
M. ESPES, David, Professeur des universités, Université de Bretagne Occidentale
Examinateurs :
M. BERRUET, Pascal, Prof. des universités, Université de Bretagne Sud
Mme. MARANGÉ, Pascale, Maître de conférences, Université de Lorraine
M. HENRY, Sébastien, Maître de conférences, IUT Lyon 1
Encadrement :
M. ZAMAÏ, Éric, Professeur des universités, INSA-LYON
Abstract :
Flexible Manufacturing Systems (FMS) aim to achieve different processes in parallel. This production strategy is made concrete with the conjoint use of flexible resources and a supervisor allocating these resources to the running processes. By design, FMSs operate in a critical environment, due to blocking allocation states or deadlock states, and uncertain, as resources can become temporarily unavailable following unexpected production events. In modern FMSs, the deployment for resources allocation and productivity enhancement of control components highly inter-connected and connected to the internet has made FMS vulnerable to cyberattacks. The origins of these cyberattacks are diverse and different attacker profiles can be defined based on their objectives, background, skills, tools and financing.
Hence, although FMSs are initially built to deal with deadlock states and resources unavailability, an expert attacker profile can be able to reach deadlock states by manipulating resources allocation decisions and resources availability. From this statement, the following research problematic arises : in FMSs uncertain environment, how can one diagnose the origin, natural or malicious, of a deadlock state and identify the attacker profile responsible for the attack ?
In answer to this problematic, three main contributions are developed. First, deadlock attacks and attacker profiles are defined and modelled in a certain environment. A deadlock attack diagnosis module is then structured from these models. Second, this module is extended to FMS uncertain environment where resource availability is considered and can be manipulated by an attacker. Third, the diagnosis module is implemented on a manufacturing platform to assess its experimental results.
Keywords:
ndustrial cybersecurity, flexible manufacturing system, discret events system, deadlock, detection, diagnosis, uncertain events