Etant membre du RUN, mes recherches portent sur les réseaux informatiques, et plus spécifiquement la découverte de topologie. Mon sujet de thèse s'intitule "Discovery and Study of the Internet Layer-2 Devices" et vise à développer des techniques d'inférence de machines de couche 2 (exemple: Ethernet switch) ainsi qu'à mesurer l'impact des mêmes machines dans l'Internet.

Ce sujet prolonge mon travail de fin d'étude "Measuring and Modeling the Internet Graph" (disponible en version PDF à cette adresse, exclusivement en anglais) qui portait sur la réalisation d'un outil de mesure utilisant de l'inférence de sous-réseaux pour reconstituer la topologie d'un réseau. Cette topologie pouvait être ensuite être convertie en graphe biparti (en utilisant un programme supplémentaire) de sorte à faciliter l'analyse des différentes propriétés des réseaux mesurés.


icon Revisiting Subnet Inference WISE-ly (full paper) - ORBi ORBi
Jean-François Grailet, Benoit Donnet
Network Traffic Measurement and Analysis Conference (TMA) 2019, Paris, 19/06/2019 - 21/06/2019

Abstract - Since the late 90’s, the Internet topology discovery has been an attractive and important research topic, leading, among others, to multiple probing and data analysis tools developed by the research community. This paper looks at the particular problem of discovering subnets (i.e., a set of devices that are located on the same connection medium and that can communicate directly with each other at the link layer).

In this paper, we first show that the use of traffic engineering policies may increase the difficulty of subnet inference. We carefully characterize those difficulties and quantify their prevalence in the wild. Next, we introduce WISE (Wide and lInear Subnet inferencE), a novel tool for subnet inference designed to deal with those issues and able to discover subnets on wide ranges of IP addresses in a linear time. Using two groundtruth networks, we demonstrate that WISE performs better than state-of-the-art tools while being competitive in terms of subnet accuracy. We also show, through large-scale measurements, that the selection of vantage point with WISE does not matter in terms of subnet accuracy. Finally, all our code (WISE, data processing, results plotting) and collected data are freely available.

icon Discovering Routers in Load-balanced Paths (extended abstract) - ORBi ORBi
Jean-François Grailet, Benoit Donnet
ACM CoNEXT 2017 Student Workshop, Séoul, 12/12/2017

Abstract - Usually, a set of Traceroute measurements collected for a large amount of target IPs contain one or several route hops at which the IP interfaces vary from one measurement to another. These variations occur even if several measurements share the same length and the same last hops. This is likely a consequence of load balancing, a traffic engineering policy which aims at sharing the load to ensure quality of service. In this paper, we consider the problem of conducting alias resolution on IP interfaces discovered via Traceroute and which are involved in load balancing. By conducting alias resolution in such a context, we want to verify if the IP interfaces involved in load balancing belong to unique routers, and more broadly, how relevant is alias resolution in this context. To do so, we use a slightly edited version of TreeNET, a topology discovery tool which relies on a tree-like structure based on Traceroute measurements to map a target domain. The upgraded TreeNET along the measurements described in this paper are both freely available online.

icon Towards a Renewed Alias Resolution with Space Search Reduction and IP Fingerprinting (full paper) - ORBi ORBi
Jean-François Grailet, Benoit Donnet
Network Traffic Measurement and Analysis Conference (TMA) 2017, Dublin, 21/06/2017 - 23/06/2017

Abstract - Since the early 2000's, the Internet Topology has been frequently described and modeled from the perspective of routers. To this end, alias resolution mechanisms have been developed in order to aggregate all IP interfaces of a router, collected with traceroute, into a single identifier. So far, many active measurement techniques have been considered, often taking advantage of specific features from network protocols. However, a lot of these methods have seen their efficiency decrease over time due to security reinforcements across the Internet.

In this paper, we introduce a generic methodology to conduct efficient and scalable alias resolution. It combines the space search reduction of TreeNET (a tool for efficiently discovering subnets) with a fingerprinting process used to assess the feasibility of several state-of-the-art alias resolution methods, using a small, fixed amount of probes. We validate our method along MIDAR on an academic groundtruth and demonstrate that our methodology can achieve similar accuracy while using less probes and discovering subnets in the process. We further evaluate our method with measurements made on PlanetLab towards several distinct ASes of varying sizes and roles in the Internet. The collected data shows that some properties of our fingerprints correlate with each other, hinting some observed profiles could be linked with equipment vendors. Both TreeNET (which implements our methodology) and our dataset are freely available.

icon TreeNET: Discovering and Connecting Subnets (full paper) - ORBi ORBi
Jean-François Grailet, Fabien Tarissan, Benoit Donnet
Traffic Monitoring and Analysis Workshop (TMA) 2016, Louvain-La-Neuve, 07/04/2016 - 08/04/2016

Abstract - Since the early 2000's, the Internet topology has been an attractive and important research topic, either for developing data collection mechanisms, and for analyzing and modeling the network. Beside traditional aspects of the Internet topology (i.e., IP interface, router, and AS levels), recent researches focused on intermediate promising visions of the topology, namely Point-of-Presence (PoP) and subnets (i.e., a set of devices that are located on the same connection medium and that can communicate directly with each other at the link layer). This paper focuses on network subnet discovery by proposing a new tool called TreeNET. One of the key aspects of TreeNET is that it builds a tree representing the way subnets are located with respect to each other. This tree allows TreeNET to obtain additional information on the network, leading to better analysis of the collected data. In this paper, we demonstrate the potential of TreeNET through the evaluation of its key algorithmic steps and the study of measurements collected from the PlanetLab testbed.