Title:
The Battle for Privacy
Abstract:
This talk is about attacking and defending IP privacy; or, in other
words, violating your boss’ privacy and protecting Wikileaks’ sources.
First, I discuss how to reverse engineer Skype to find out the IP
address of a targeted person.  This could be exploited for many
malicious purposes such as finding out whether your boss is
downloading copyrighted-porn material.  This privacy attack can be
scaled to track, for example, all users working for a large
corporation such the Universal music group.  Furthermore, the attack
can be launched from the comfort of one’s home, without any legal
power, and without the targeted users knowing.  (The story goes that
the first author of this research used this attack to track his
Ph.D. committee.)
Second, I present a system for efficiently defending against IP-based
privacy attacks, even when the attacker is a powerful government such
as the USA. I motivate our novel design by showing that onion routing,
the reference design for IP anonymity, has fundamental scaling issues
when defending against an attacker able to analyze packet logs.  On
the other hand, I show that our design scales with the number of users
while being more efficient than existing solutions.  By combining
anonymity and efficiency, we expect our system to appeal to regular
Internet users, such as visitors of the late MegaUpload, and
whistleblowers alike.
Bio:
Stevens Le Blond holds a M.Sc. from Vrije Universiteit, Amsterdam and
a Ph.D. from INRIA Sophia Antipolis. Stevens seeks to tackle
high-risk/high-impact research problems in the system and networking
areas. At the moment, he believes that many such problems lie in
online privacy.  After enjoying sun bathing in Nice, Stevens is now
enjoying actual bathing in Germany where he’s a postdoc in Paul
Francis’ group at MPI-SWS.

Title:
——-
Resource provisioning for complex Web applications

Abstract:
————-
Dynamic resource provisioning aims at maintaining the end-to-end response time of a web application within a pre-defined range (Service Level Objective, SLO). Provisioning resources for applications composed of multiple services remains a challenge. When the SLO is violated, one must decide /which/ service(s) should be re-provisioned for optimal effect. We propose to assign an SLO only to the front-end service. Other services are not given any particular response time objectives. Services are autonomously responsible for their own provisioning operations and collaboratively negotiate performance objectives with each other to decide the provisioning service(s). After presenting the resource provisioning techniques themselves, I will discuss their application and implementation in the context of the ConPaaS runtime environment for elastic Cloud applications.

Bio:
——
Guillaume Pierre is an associate professor at VU University Amsterdam. His research focuses on the management of very large-scale distributed systems. He particularly studied Web applications as a good example of demanding large-scale systems and Cloud computing platforms. His research addresses a variety of questions such as how to make applications scale, how to control their non-functional properties, and how to design large-scale decentralized infrastructures where they can be deployed.

Tolérer les fautes transitoires, permanentes et intermittentes

Lorsque la taille d’un système réparti devient importante ou lorsque ce système est déployé dans un environnement non contrôlé, la probabilité que certains éléments du système subissent des fautes (panne, corruption de mémoire, piratage, …) devient non négligeable. Ces fautes peuvent être classifiées en fonction de leur durée, de leur étendue et de leur nature. Dans ce séminaire, nous nous intéressons aux systèmes répartis capables de tolérer simultanément plusieurs types de fautes à travers l’étude de trois problèmes fondamentaux. Nous présentons ainsi un protocole réparti simulant un registre atomique mono-écrivan multi-lecteurs en présence de fautes transitoires et de fautes permanentes de type crash. Ce protocole repose sur deux outils ré-utilisables : un protocole de communication et un système d’estampillage borné. Ensuite, nous proposons une étude de la synchronisation faible d’horloges logiques en présence de fautes transitoires et de fautes intermittentes Byzantines. Finalement, nous définissons trois nouveaux concepts de tolérance pour les systèmes répartis sujets à des fautes transitoires et des fautes intermittentes Byzantines. Nous donnons un protocole de construction d’une vaste classe d’arbres couvrants optimal selon ces trois concepts.

Bern, Switzerland – 10 April 2012.

http://www.cambridgeplus.net/SNS12/

Online social networks (e.g., Facebook, Linkedin, Twitter, etc.) are among the most popular sites on the Web and are still growing rapidly. These systems have their own characteristics and challenges. However, the popularity of these new systems is not reflected in current systems research. There is no major venue or conference dedicated to system support for social networks. Little research has examined these systems or addressed their needs, for instance fundamental problems like managing and online querying of large social graphs are still mostly unchartered territories.
The workshop addresses this situation by providing a forum for researchers to discuss the systems issues associated with online social networks. The workshop nicely complements the topics of the main EuroSys conference, since the research issues with online social networks span the topics of distributed computing, databases, and storage systems as well as security and privacy. The ultimate goal of this workshop is to bring together the researchers and practitioners who are active in this emerging area.

« Newer Posts - Older Posts »