abstract: The field of cyber-defenses has played catch-up in the cat-and-mouse game of finding vulnerabilities followed by the invention of patches to defend against them. With the complexity and scale of modern-day software, it is difficult to ensure that all known vulnerabilities are patched; moreover, the attacker, with reconnaissance on their side, will eventually discover and leverage them. To take away the attacker's inherent advantage of reconnaissance, researchers have proposed the notion of proactive defenses such as Moving Target Defense (MTD) in cyber-security. In this thesis, I make three key contributions that help to improve the effectiveness of MTD. First, I argue that naive movement strategies for MTD systems, designed based on intuition, are detrimental to both security and performance. To answer the question of how to move, I (1) model MTD as a leader-follower game and formally characterize the notion of optimal movement strategies, (2) leverage expert-curated public data and formal representation methods used in cyber-security to obtain parameters of the game, and (3) propose optimization methods to infer strategies at Strong Stackelberg Equilibrium, addressing issues pertaining to scalability and switching costs. Second, when one cannot readily obtain the parameters of the game-theoretic model but can interact with a system, I propose a novel multi-agent reinforcement learning approach that finds the optimal movement strategy. Third, I investigate the novel use of MTD in three domains-- cyber-deception, machine learning, and critical infrastructure networks. I show that the question of what to move poses non-trivial challenges in these domains. To address them, I propose methods for patch-set selection in the deployment of honey-patches, characterize the notion of differential immunity in deep neural networks, and develop optimization problems that guarantee differential immunity for dynamic sensor placement in power-networks. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2020

L’objectif de cette thèse est de développer et étudier des modèles mathématiques d’échanges économiques, basés sur la flexibilité de la demande, entre fournisseurs et consommateurs d’électricité. D’une part, des fournisseurs d’électricité offrent des prix dépendant de l’heure de consommation. D’autre part, des consommateurs adaptent leur usage, minimisant leur facture et le désagrément lié aux changements de consommation induits. La structure de ces problèmes correspond à des problèmes d’optimisation bi-niveau. Trois types de modèles sont étudiés. Tout d’abord, l’interaction entre un fournisseur et un opérateur de smart grid est modélisée par un problème à un seul meneur et un seul suiveur. Pour cette première approche, le niveau de détails du suiveur est particulièrement élevé, et inclut notamment une gestion stochastique de la production distribuée. La meilleure réponse d’un fournisseur dans un modèle à plusieurs meneurs et plusieurs suiveurs fait l’objet de la seconde partie de la thèse. Celle-ci intègre aussi la possibilité d’avoir des agrégateurs comme suiveurs. Deux nouvelles méthodes de résolution reposant sur la sélection d’équilibres de Nash entre suiveurs sont proposées. Enfin, dans une troisième et dernière partie, on se focalise sur la recherche d’équilibres non coopératifs pour ce modèle à plusieurs meneurs et plusieurs suiveurs.Tous les problèmes abordés dans cette thèse le sont non seulement d’un point de vue théorique, mais également d’un point de vue numérique / The objective of this thesis is to develop and study mathematical models of economical exchanges between energy suppliers and consumers, using demand-side management. On one hand, the suppliers offer time-of-use electricity prices. On the other hand, energy consumers decide on their energy demand schedule, minimizing their electricity bill and the inconvenience due to schedule changes. This problem structure gives rise to bilevel optimization problems.Three kinds of models are studied. First, single-leader single-follower problems modeling the interaction between an energy supplier and a smart grid operator. In this first approach, the level of details is very high on the follower’s side, and notably includes a stochastic treatment of distributed generation. Second, a multi-leader multi-follower problem is studied from the point of view of the best response of one of the suppliers. Aggregators are included in the lower level. Two new resolution methods based on a selection of Nash equilibriums at the lower level are proposed. In the third and final part, the focus is on the evaluation of noncooperative equilibriums for this multi-leader multi-follower problem.All the problems have been studied both from a theoretical and numerical point of view.