Success Stories 2011
During year 2011, a lot of research activities have been performed within ICTEAM. More than 360 articles were published in international journals and conferences. Twenty-nine PhD degrees were awarded within the institute. To illustrate the quality of the research carried out within ICTEAM, we describe in this web page a few success stories that show both the breadth and the depth of our research.
The vision of the Internet-of-Things (IoT) as the core of the Web3.0 calls for the deployment of up to one trillion wireless sensor nodes. A study of the ICTEAM in collaboration with UC Berkeley and P.E. International shows that this deployment comes at a cost: the carbon footprint of their integrated circuits. To minimize the potential threat for our environment, a team of ICTEAM researchers (J. De Vos, C. Hocquet, F. Botman and F. Durvaux) lead by Dr David Bol designed the first low-carbon microcontroller for IoT wireless sensor nodes, codenamed SleepWalker.
SleepWalker is a full System-on-Chip based on the popular MSP430 architecture from Texas Instruments and fabricated in the 65nm CMOS technology from STMicroelectronics. It embeds computation, memory, data acquisition and power management functionalities on a piece of silicon smaller than 1mm², which ensures a low carbon footprint for chip fabrication. The researchers leveraged the unique design techniques proposed in the 2008 Ph.D thesis of David Bol to allow robust and fast operation at a supply voltage as low as 0.4V. This let SleepWalker achieve a 20x reduction of power consumption when compared to the commercial MSP430 from Texas Instruments. Moreover, this is the first time that a circuit operated at such a low voltage achieves speed performances of 25MHz, which makes SleepWalker 50x faster than the 0.5V counterpart from the Massachusetts Institute of Technology. In other words, SleepWalker walks when others sleep.
SleepWalker was presented at the prestigious IEEE International Solid-State Conference (ISSCC) in February 2012 at San Francisco and received a lot of attention from the industrial and academic audience. Since then, the authors were invited to present their work in the IEEE Journal of Solid-State Circuits and SleepWalker has been pointed out by the Guardian Angel consortium as the reference microcontroller in the roadmaps of their Flagship European project. With SleepWalker, the team hopes to raise awareness of circuit designers and decision makers in the field of semiconductors on the environmental impact of their products.
The Louvain method developed in ICTEAM is a simple and remarkably efficient method for detecting communities in very large networks. The method was introduced by the team of Prof. Vincent Blondel in 2008 and has immediatly attracted wide interest. It has since then been used for problems as diverse as describing the hierarchy of the human brain network, detecting emerging scientific specialties, discriminating between types of words in linguistic text data, detecting computer viruses through their patterns of dissemination, forecasting political events, uncovering major trading changes during the recent credit crisis, or monitoring wildlife.
The method based on an original idea by master student Etienne Lefebvre allows to routinely analyze networks with billions of nodes and connections on a standard laptop. The method greedily unveils communities and allows to zoom within communities to discover the hierarchical sub-community structure. It is today one of the most widely used method for detecting communities in large networks.
The initial article describing the method attracts one new citation every second day and the method has been implemented in numerous softwares, including in Gephi, NetworkX and in the Brain Connectivity Toolbox. The Louvain method was recently selected by LinkedIn as the method of choice for the visualization application InMaps. According to LinkedIn’s developpers « the methods gives excellent results and is really really fast ». LinkedIn now offers the network visualization application to its 150 million customers.
The UCL machine learning group has developed an expertise in bio-medical data analysis, typically for the discovery of bio-markers from DNA microarray gene expression measurements. The research team of Prof. P. Dupont has fruitful collaborations with medical researchers at the Institute of Experimental and Clinical Research from the UCL. Research projects include the early diagnosis of rheumatoid arthritis (Prof. Bernard Lauwerys), the discovery of molecular signatures representative of intermittent hypoxia in tumors (Prof. Olivier Feron) and the determination of the risk criteria for allergy in babies (Prof. Etienne Sokal, Prof. Françoise Smets). A collaboration with GlaxoSmithKline Biologicals also lead to the discovery of markers predictive of the response of cancer patients to an immunotherapy.
The work of Thibault Helleputte within this group has already been awarded on several occasions. His master thesis, entitled "Use and Analysis of Microarrays in a Liver Cell Transplantation Therapy", received the Baudouin Elleboudt Award (best computer science master thesis) in 2006. His PhD Thesis, entitled "Inductive Biases for Stable Feature Selection in High Dimensional Spaces: Applications to Gene Profiling and Diagnosis from DNA Microarrays", received the IBM Belgium / F.R.S.-FNRS 2011 Award for the Best PhD Thesis in Computer Science and Applications. This work offers novel contributions to the problem of selecting discriminative variables in high dimensions from very sparse data.
After completing his PhD, Dr. Thibault Helleputte has been funded by a First Spin-off project of the Walloon Region to promote the transfer of fundamental research results to an industrial activity in biotechnology and the health domain.
The DNAlytics spin-off project has been awarded as the best health project during the Mind and Market forum in April 2011. DNAlytics also received the first prize - Innovation Award - of the 4th edition of the Microsoft Innovation Center Boostcamp in December 2011.
The ERC Starting Independent Researcher Grant is an important award (worth up to one and a half million euros for a five-year period), for which thousands of candidates apply each year. ICTEAM member François-Xavier Standaert was among the 12% of successful applicants in 2011, for a project on Cryptographic Algorithms and Secure Hardware. The grant envisions new strategies to improve the security of cryptographic implementations against so-called physical attacks, e.g. side-channel analysis in which an adversary takes advantage of information leakages (such as the power consumption of a smart card, in order to recover sensitive data. By allowing to circumvent the models in which standard security proofs are obtained, these attacks exhibit a gap between the mathematical abstractions of modern cryptography and the concrete peculiarities of actual electronic circuits. Hence, by considering physical and algorithmic issues in a unified way, the CRASH project is expected to get rid of the incompatibilities between the different models that can be used to explain the information leakage in cryptographic implementations. In other words, it aims to formalize side-channel attacks and to efficiently prevent them, in order to develop concrete basements for the next generation of cryptographic algorithms and secure hardware.
For this purpose, the precise understanding of physical attacks is a preliminary. It requires and inter-disciplinary approach, including the investigation of measurements setups, the statistical analysis of leakages, their information theoretic evaluation and exploitation with sound computational strategies, together with the modeling of these different aspects in a comprehensive mathematical framework for proving (or arguing about) the security. These different questions already triggered various collaborations within the ICTEAM institute, e.g. with Prof. Olivier Pereira for provable security issues and leakage-resilience (see, e.g. the paper published at ACM CCS 2010), and Prof. Denis Flandre for studying the connection of physical security issues with technology scaling (see, e.g. the paper published at Eurocrypt 2011).
The Transmission Control Protocol (TCP) is one of the most important protocols on the Internet today. TCP is a transport protocol that ensures the reliable delivery of the packets through the network. TCP is able to detect packet losses and deals with them by retransmitting the lost packets and adjusting its transmission rate to adapt to congestion. When TCP was designed, in the 1970s, computers were attached to the network via a single interface. In order to minimize the transmission overhead, the designers of TCP chose to strongly couple TCP with the Internet Protocol (IP). Since then, a TCP connection is bound to the two IP addresses of the endpoints of the connection. This design was perfectly valid for an Internet composed of fixed hosts. In this environment, the most common failures were the link failures inside the network. These failures are handled by IP thanks to the routing protocols that dynamically reroute packets when the topology changes.
During the last decade, the Internet has significantly changed. The Internet connects a growing population and more and more Internet hosts are equipped with multiple interfaces. Servers often use multiple Ethernet interfaces for redundancy and performance reasons. On the other hand, smartphones, laptops and tablets are now equipped with Ethernet, 3G and WiFi interfaces. End users expect that thanks to these multiple interfaces, applications should be able to continue to operate even when they move (and thus change their IP address). With the original TCP design, this is unfortunately not possible because established TCP connections must be torn down as soon as one of the IP address of the endpoints changes.
Various researchers have worked on this problem during the last decade. Some have proposed to change the IP protocol in order to hide the changes in IP addresses to TCP. Others have proposed to introduce new transport protocols. As of today, these solutions are still not widely deployed. A different approach was chosen within the FP7 Trilogy project. Instead of designing a new protocol, this project chose to develop backward compatible extensions to TCP to allow it to operate efficiently in this environment, called Multipath TCP. In a few words, Multipath TCP allows two hosts to establish a TCP connection to reliably exchange data. Once the TCP connection has been established, these two hosts are allowed to use other interfaces and IP addresses, during the lifetime of the connection. Multipath TCP enables servers to use multiple interfaces at the same time to increase throughput. It also enables smartphones to start a connection over a WiFi interface and switch to the 3G interface when it moves without affecting the applications. The video below shows visually the operation of Multipath TCP on a laptop with Ethernet, 3G and WiFi interfaces. The laptop connects to a server and uses XWindow over ssh to export a graphical application that regularly refreshes the screen over a Multipath TCP connection. During the lifetime of the connection, we disable the Ethernet, 3G and WiFi interfaces without significantly affecting the operation of MultipThere are various other use cases for Multipath TCP. With regular TCP, the connection would break as soon as the Ethernet interface stops. Multipath TCP intelligently balances the load over the different interfaces.
Several ICTEAM researchers have played a key role in the design of the Multipath TCP protocol. Sébastien Barré participated to the design of the initial architecture and designed the first Multipath TCP implementation in the Linux kernel. Olivier Bonaventure is one of the authors of the Multipath TCP protocol. Christoph Paasch, Gregory Detal and Fabien Duchêne have significantly improved the Multipath TCP implementation in the Linux kernel. This is currently the reference implementation of the Multipath TCP protocol. It received the community award at USENIX NSDI 2012, one of the leading networking conferences. This work has been funded by the European Commission within the FP7 Trilogy and CHANGE projects and a research grant from Google. It will continue within the forthcoming FP7 Trilogy 2 project.
For additional information :
The International Conference on Image Processing, sponsored by the IEEE Signal Processing Society, was organized in 2011 by Prof. Benoît Macq (UCL/ICTEAM) and Prof. Peter Schelkens (VUB). This exceptional conference, 18th in a series that has been held annually since 1994, took place on September 11-14, 2011 at the Square Brussels Meeting Center and brought together leading engineers and scientists in image processing from around the world. This event attracted more than 1,200 participants.
ICIP is the premier forum for the presentation of technological advances and research results in the fields of theoretical, experimental, and applied image and video processing. Research frontiers in fields ranging from traditional image processing applications to evolving multimedia and video technologies are regularly advanced by results first reported in ICIP technical sessions.
For ICIP 2011, as well as the general topics commonly used, emphasis was also placed on a series of special Belgian flavored themes, namely astronomy, cultural heritage and human media interaction.
General ICIP 2011 themes included :
These topics were synchronized with the invitation of renowned international researchers for special keynote lectures, namely, Prof. Ingrid Daubechies (Duke Univ., USA), Prof. Brian Wandell (Stanford University, USA), Prof. Jean-Luc Starck (CEA-Saclay, France), Prof. Pierre Moulin (University of Illinois, USA) and Prof. Ton Kalker (Huawei Technologies).
The program also included Tutorials, Special Sessions (i.e. oral technical sessions with 8 invited papers), Lectures and Poster Sessions. A Best Paper Award and Student Best Paper Award were also granted.
Second prize for SIAM 'Math Matters, Apply it!' contest goes to ICTEAM PhD students
The second prize of the 'SIAM Math Matters, Apply It!' 2011 contest was awarded to Nicolas Boumal (ICTEAM), Romain Hollanders (ICTEAM) and Karim Slaoui (iMMC) for their poster about the role of Independent Component Analysis in fetal ECG. The first place was awarded to a student at Ohio University (USA) for a poster about social networks.
The 'SIAM Math Matters, Apply It!' contest is a worldwide competition open to all students and organized by SIAM, the Society for Industrial and Applied Mathematics. Participants are to submit short descriptions in layman's terms exemplifying the central role of mathematics in a modern application that affects people's lives. Winning submissions are transformed into posters to promote mathematical education.
| 21/02/2013 |