The JFLI Internship Day 2016 will be held on:
Date: July 13th, 2016
Place: National Institute for Informatics (NII) – Room 1902/1903 – 19th floor
Time: 14:00 – 17:00
Program: Seven presentations by students about their research internship
14:00-14:10: Welcome: What is a Research Internship?, Kae Nemoto and Phong Nguyen
14:10-15:30: Session on Cybersecurity
14:10-14:30: Arthur Carcano (ENS and IIJ), Traffic aggregation: tools, performances and assessment
Abstract: « Who are we ? Why are we ? Why are my network performances so bad ? » Questions like this come haunting thousands of sysadmins across the world everyday. Indeed, traffic analysis can quickly become a daunting task, shall one undertake it without the appropriate tools. One of this tools is called agurim, and I have been working on assessing its performances for the past four months. Hence, this talk will try to summarize what are the challenges, ideas, and issues of the assessment of traffic aggregation tools. »
14:30-14:50: Gwendoline Chouasne-Guillon (ENS and IIJ), Detecting BGP Anomalies
Abstract: BGP (Border Gateway Protocol) is the routing protocol of the Internet. However, BGP misconfigurations still frequently occur with a world-wide impact. We present a way to detect BGP anomalies based on received BGP messages and graph theory.
14:50-15:10: Richard Garnier (ENSIMAG and NII), Privacy protection technologies: analysis and comparison
Abstract: In this context, companies track users to provide them targeted ads. This phenomenon causes privacy concerns. In response, several privacy protection technologies (such as browser add-ons) have been developed. This study assesses the effectiveness of existing privacy protection technologies. To the best of our knowledge, this is the first study that extensively compare a wide set of privacy protection technologies. This study has 3 main findings: 1) When available, localized blocking lists improve add-ons performance in regards to default blocking lists. 2) The Request Policy Continued and NoScript add-ons exhibit good result but impact websites quality. 3) The Ghostery add-on provides slightly lower performance, but respects websites’ functionalities.
15:10-15:30: Yixin Shen (Ecole Polytechnique and the University of Tokyo), Recovering a Hidden Number with the Fast Fourier Transform
Abstract: In 2000, Bleichenbacher discovered a clever attack against one of the most widely used cryptographic algorithms: the DSA digital signature standard. His attack took advantage of a subtle bias in the DSA pseudo-random number generator. At the heart of the attack is a heuristic FFT-based algorithm to solve the so-called hidden number problem. We study whether this algorithm can be rigorously analyzed or not, using several points of view.
15:30-16:00: Coffee Break
16:00-17:00: Session on Session on Graphs and Networks
16:00-16:20: Luca Wehrstedt (ENS and NII), Colorability of odd-K5-minor-free graphs
Abstract:H is an odd-minor of a graph G if it can be obtained from G by repeatedly removing vertices and edges and by simultaneously contracting all edges that cross a cut. Odd-K5-minor-free graphs don’t contain K5 as an odd minor. They include planar graphs. My project aims at generalizing the four-color theorem to this class, by proving that it is 4-colorable.
16:20-16:40: Vivek Claver (Ecole Polytechnique and NII), Influence in citation networks, a multiplex flow approach
Abstract: Researchers are evaluated from the impact and the popularity of their work. This is often measured by their number of citations, H-Index, and other similar measures. However measuring the real impact of a published work still remains a challenging task, and often may depend on the domain and the popularity of the author. Some works may well be seminal in their discovery, but not in their diffusion. Interested by this process of creation and diffusion of knowledge, we propose in this work an exploration of the citation networks from the perspective of flows. We also contribute by bringing a novel approach through a multiplex interpretation of flows in citation networks.
16:40-17:00: David Delzant (Ecole Polytechnique and NII), Studying political leaning through graph signal processing
Abstract: « Tell me who are you friends and I’ll tell you who your are », this quote has never been so true since the recent years, with social network analysis becoming very popular. In most developed countries, the world of politics is often polarized, with political positions placed along a more or less aligned space, often designated from Left to Right. Recent progress in video analysis allow us to track well known politicians in a large TV archive. Since TV shows the politico-media scene of a country, it reports the activity of political figures, from which we can derive social networks. In this work, we want to investigate how the public activity of these politicians reflects their political leaning, and if we could attribute specific leaning to politicians based on their public image behavior, using graph signal processing models.