Matthieu Latapy, Elie Rotenberg, Christophe Crespelle, Fabien Tarissan

13th IFIP International Conference on Networking – Networking 2014, 2014, Trondheim, Norway. IEEE, pp.1-9

Most current models of the internet rely on knowledge of the degree distribution of its core routers, which plays a key role for simulation purposes. In practice, this distribution is usually observed directly on maps known to be partial, biased and erroneous. This raises serious concerns on the true knowledge one may have of this key property. Here, we design an original measurement approach targeting reliable estimation of the degree distribution of core routers, without resorting to any map. It consists in sampling random core routers and precisely estimate their degree thanks to probes sent from many distributed monitors. We run and assess a large-scale measurement following this approach, carefully controlling and correcting bias and errors encountered in practice. The estimate we obtain is much more reliable than previous knowledge, and it shows that the true degree distribution is very different from all current assumptions.

Download

Fabien Tarissan, Elie Rotenberg, Matthieu Latapy, Christophe Crespelle

IEEE 22nd International Symposium on Modeling Analysis and Simulation of Computer and Telecomunication Systems (MASCOTS’14), At Paris, France

The classical approach for Internet topology measurement consists in distributively collecting as much data as possible and merging it into one single piece of topology on which are conducted subsequent analysis. Although this approach may seem reasonable, in most cases network measurements performed in this way suffer from some or all of the following limitations: they give only partial views of the networks under concern, these views may be intrinsically biased, and they contain erroneous data due to the measurement tools. Here we present a new tool, named UDP Ping , that relies on a very different approach for the measurement of the Internet topology. Its basic principle is to measure the interface of a given target directed toward a monitor which sends the measurement probe. We demonstrate how to use it to deploy real world-wide measurements that provide reliable (i.e. bias and error free) knowledge of the Internet topology, namely the degree distribution of routers in the core Internet in our example.

Download

Alice Albano, Jean-Loup Guillaume, and Bénédicte Le Grand

Proceedings of the 8th IEEE International Conference on Research Challenges in Information Science (RCIS 2014)

The analysis of social networks is a challenging research area, in particular because of their dynamic features. In this paper, we study such evolving graphs through the evolution of their community structure. More specifically, we build on existing approaches for the identification of stable communities over time. This paper presents two contributions. We first propose a new way to compute such stable communities, using a different time scale, called intrinsic time. This intrinsic time is related to the dynamics of the graph (e.g., in terms of link appearance or disappearance) and independent from traditional (extrinsic) time units, like the second. We then show how visualization both at intrinsic and extrinsic time scales can help validating and interpreting the obtained communities. Our results are illustrated on a social network made of contacts among the participants of the 2006 edition of the Infocom conference.

Download

Lionel Tabourier, Anne-Sophie Libert, and Renaud Lambiotte

2014, DyNakII, 2nd International Workshop on Dynamic Networks and Knowledge Discovery (PKDD 2014 workshop)

In this work, we consider the issue of unveiling unknown links in a social network, one of the difficulties of this problem being the small number of unobserved links in comparison of the total number of pairs of nodes. We define a simple supervised learning-to-rank framework, called RankMerging, which aims at combining information provided by various unsupervised rankings. As an illustration, we apply the method to the case of a cell phone service provider, which uses the network among its contractors as a learning set to discover links existing among users of its competitors. We show that our method substantially improves the performance of unsupervised metrics of classification. Finally, we discuss how it can be used with additional sources of data, including temporal or semantic information

Download

Qinna Wang

2014 International Conference on Data Science and Advanced Analytics (DSAA2014)

We tackle the problem of predicting future links in dynamic networks. For this, we work with the Debian Mailing Lists. In this dataset, a user can post a question to the debian list and other users can reply it by email forming a thread. We show that the number of threads shared in the past between users is a better feature to predict future email exchanges than classical features, like the number of common neighbors. We also show that the structure of a thread do not match the traditional definition of a community, particularly a thread does not have many triangles and has many outgoing connections. While the number of shared (detected) communities is also a better feature to predict future email exchanges than traditional features, is not as good as the number of shared threads. We believe our work should raise interests in characterizing and detecting thread-like structures in dynamic networks.

Download

Maximilien Danisch, Nicolas Dugué and Anthony Perez

2014 International Conference on Behavioral, Economic, and Socio-Cultural Computing (BESC 2014)

Influence on Twitter has drawn a lot of attention these past few years since this microblogging service is used to share, seek or debate about any kind of information. Several tools providing so-called influential scores have thus been proposed. However, the algorithms behind them are kept secret and it is not clear how they consider influence. Yet, many users rely on such tools to evaluate and even try to improve their influence in the Twitter network. In a recent work, it has been shown that automatic accounts can obtain high influential scores with no intuitive reason. Extending and completing this work, we show that such measures fail at distinguishing so-called social capitalists from real, truthful users. This enlights the fact that actual scores do not seem to consider the way followers and interactions are obtained on the network. To overcome this issue, we define a classifier that discriminates social capitalists from truthful users. To that aim, we crawled the Twitter network to gather examples of certified social capitalists and regular users and obtained features related to the profile and behavior of each user. We then use such a classifier to balance Klout’s score to adjust influential scores. We also developed an application that allows using our classifier online. We believe our work should raise the question of the legitimacy of influence on Twitter, and lead to significant improvements in the way it is measured.

Download

Maximilien Danisch, Jean-Loup Guillaume and Bénédicte Le Grand

2014 International Conference on Data Science and Advanced Analytics (DSAA2014)

In large-scale online complex networks (Wikipedia, Facebook, Twitter, etc.) finding nodes related to a specific topic is a strategic research subject. This article focuses on two central notions in this context: communities (groups of highly connected nodes) and proximity measures (indicating whether nodes are topologically close). We propose a parameterized proximity measure which, given a set of nodes belonging to a community, learns the optimal parameters and identifies the other nodes of this community, called multi-ego-centered community as it is centered on a set of nodes. We validate our results on a large dataset of categorized Wikipedia pages and on benchmarks, we also show that our approach performs better than existing ones. Our main contributions are (i) a new ergonomic parametrized proximity measure, (ii) the automatic tuning of the proximity’s parameters and (iii) the unsupervised detection of community boundaries.

Download

Darko Obradovi, Maximilien Danisch

The 6th International Conference on Computational Aspects of Social Networks (CASON2014)

This paper intends to extend the possibilites available to researchers for the evaluation of directed networks with the use of randomly generated graphs. The direct generation of a simple network with a prescribed degree sequence still seems to be an open issue, since the prominent configuration model usually does not realise the degree distribution exactly. We propose such an algorithm using a heuristic for node prioritisation. We demonstrate that the algorithm samples approximately uniformly. In comparison to the switching Markov Chain Monte Carlo algorithms, the direct generation of edges allows an easy modification of the linking behaviour in the random graph, introducing for example degree correlations, mixing patterns or community structure. That way, more specific random graphs can be generated (non-uniformly) in order to test hypotheses on the question, whether specific network features are due to a specific linking behaviour only. Or it can be used to generate series of synthetic benchmark networks with a specific community structure, including hierarchies and overlaps.

Download

Raphaël Fournier, Maximilien Danisch

IEEE International Conference on Research Challenges in Information Science 2014 (RCIS2014)

Peer-to-peer (P2P) networks are popular to exchange large volumes of data through the Internet. Pedophile activity is a very important topic for our society and some< works have recently attempted to gauge the extent of pedophile exchanges on P2P networks. A key issue is to obtain an efficient detection tool, which may decide if a sequence of keywords is related to the topic or not. We propose to use social network analysis in a large dataset from a P2P network to improve a state-of-the-art filter for pedophile queries. We obtain queries and thus combinations of words which are not tagged by the filter but should be. We also perform some experiments to explore if the original four categories of paedophile queries were to be found by topological measures only.

Download