Research Activities

My research lies in strategic decision-making and its consequences when the actions of each agent have an impact on the others. This research is at the crossroad between the disciplines of mathematics, computer science and economics.

Research themes

Among the research questions I am investigating, we can distinguish four main orientations, summarized below.

Attack and defense strategies in networks

The objective is to model a strategic interaction situation where two players face each other: one player, the attacker, is attempting to disconnect the network built by the other player, the defender. At the end of the attack, each players' payoff depends on the connectivity of the residual network, as well as the costs associated with creating and attacking links with different levels of protection. We study emerging network topologies.
This type of question can cover very different realities and applications, both in economics (e.g. protecting a production line), and in cybersecurity (maintaining a communication system or access to a server).
In some situations, the attack is not "all or nothing". For example, in the case of communication jamming in wireless networks, the link is not necessarily broken, but its operation may be degraded or its capacity reduced. The question then becomes: **What overall quality-of-service (QoS) can the system still guarantee?

Mechanisms of influence in social networks

An emerging form of strategic decision-making, typical of social networks, is that in which agents seek to modify the structure of the social influence network through the choice of messages they send to it.
Thus, to increase their number of subscribers(1), social influencers may choose posts (messages) according to their own anticipation of the beliefs of others. Conversely, a popular influencer has an impact on the overall belief present on the social media.
Therefore, the goal is to study the optimal strategies of influencers and the dynamics of the network structures resulting from such strategic disclosures. Conversely, the network administrator or public authority should measure the relative weights of influencers in the evolution of global belief, and act strategically so as to protect users from the creation of information bubbles. In other words: influence the influencers!

1. Formely, its connectivity degree in the interaction graph.

The paradoxes caused by non-cooperative decisions

When different agents do not coordinate their actions and individually (selfishly) optimise their own interests (what is called non-cooperative strategies), the resulting equilibria are generally sub-optimal. A consequence is the possible occurrence of paradoxes, phenomena in which adding resources to the system causes a deterioration in the performance of all players.
The analysis of situations that give rise to paradoxes is essential in the deployment of large distributed systems, especially when the operator cannot control individual behaviour. This involves both identifying the situations in which paradoxes occur and quantifying their impact.
Paradoxes have already been identified and studied in a variety of networking systems (road, communication, queuing, etc).
The addition of resources is not the only cause of paradoxes in systems of uncoordinated actors: contrary to the popular adage "there is strength in numbers", the formation of coalitions or an increase in the level of information can have detrimental effects. Even more surprisingly, in some cases, one or more altruistic players can inadvertently damage the well-being of all!

Fair sharing of resources in networks

Common language often associates fair share with equal share. In fact, fairness takes into account the diversity of needs and the relative satisfaction perceived through the allocation of resources. In economics, several families of formal (mathematical) definitions of fair share have been proposed as the result of a bargaining process (1).
In networks, two families of questions arise. First, what are the relevant definitions of fairness in different application contexts (electricity transport, distributed computing platforms, wired and wireless communications in particular)? Second, how should they be implemented in systems? A distinction can be made between implementations

• at system level: under which conditions can networks be designed to ensure fair sharing of resources in the presence of selfish users?
• at user level: if only for network throughput, our machines are constantly carrying out (via computer protocols) virtual bids and other automatic negotiations without us even being aware of it!

1. That is to say a mechanism of offers and counter-offers made according to certain pre-defined rules.

Supervision of students and young researchers

Here is the list of doctoral students and post-doctoral fellows that I currently supervise (highlighted) or have supervised:

• Post-doctorates:

  • Ioannis Stiakogiannakis (2015), "Orchestration of distributed learning mechanisms for resource allocation in mobile networks" (funding: ANR project NETLEARN), with Panayotis Mertikopoulos.
  • Nof Abuzainab (2014-2015), "Fairness mechanisms and coalition formation in wireless networks (cognitive radio and millimeter waves)" (funding: Google grant).

• Doctoral students:

  • Jérémy Petithomme (Université Grenoble Alpes, since 2023), “Graph dynamics: an application to social networks” (funding: AEx project MID-ToRS).
  • Mandar Datar (Avignon université, 2018-2022), "Resource allocation and pricing for network slicing in 5G networks" (funding: Inria-Nokia Bell Labs joint laboratory), co-advising with Eitan Altman. Manuscript
  • Pierre Coucheney (Université Grenoble Alpes, 2008-2011), "Self-optimization of wireless networks" (funding: joint Inria-Alcatel Lucent laboratory), co-advising with Bruno Gaujal. PhD Defense - Manuscript (French only)
  • Rémi Bertin (Université Grenoble Alpes, 2007-2010), "Collaborative Incentives in Peer-to-Peer Networks and Application to Collaborative Computing" (funding: ANR DOCCA), co-advising with Arnaud Legrand.
  • Atsushi Inoie (Tsukuba University (Japan), 2003-2006), "Studies on Optimal Control Problems in Communication Networks with Multiple Users", unofficial co-supervision. Official advisor: Prof. Hisao Kameda. Manuscript.

International visits and collaborations

• Technion (Israel) (Oct. 2019-Jan. 2020): collaboration with Ariel Orda (invitation). We worked on a strategic availability announcement problem in a logistics system.
• New York University (USA) (Oct.-Nov. 2016 and Mar.-Apr. 2017): collaboration with Quanyan Zhu (invitation). We worked on optimal strategies in cybersecurity.
• Calcutta University (India) (Dec. 2015-Jan. 2016): collaboration with Ujjwal Maulik and Sanghamitra Bandyopadhyay (project BiDee). We have worked on new paradigms for strategic electricity exchanges and the use of exchange markets.
• Tsukuba and Kyoto Universities (Japan) (Feb.-Mar. 2010): collaboration with Li Jie and Hisao Kameda (funding: JSPS Bridge program). We worked on the study of paradoxes resulting from the inefficiency of Nash equilibria in networks (with Prof. Kameda) and routing in new wireless networks (with Prof. Li).
• Universidade Federal do Rio Grande do Sul (UFRGS) in Porto Alegre (Brazil) (Jul.-Aug. 2007): collaboration with Nicolas Maillard’s team (1). We studied the relevance and applicability of game theory tools for high performance computing.
• IBM research (Thomas J. Watson Research Center) (USA) (Jun.-Aug. 2003): collaboration with Laura Wynter and Parijat Dube on a pricing model for services with variable qualities (QoS) in a competitive market.

1. This visit was part of the partnership strategies between the MOAIS and MESCAL teams at Inria and the UFRGS on high performance computing.