MSR Inria Joint Lab
Microsoft Research and Inria renew their research partnership
Jean-Philippe Courtois, President of Microsoft International, and Michel Cosnard, President and CEO of Inria, announced today the renewal of their partnership on the Microsoft Research-Inria Joint Centre for another four years.
In beginning a new ambitious cycle for this worldwide public/private research centre, Microsoft Research and Inria reinforce their commitment to basic research and innovation. This new agreement will enable our team of researchers, who work in the heart of the future Paris-Saclay University, to open their research scope to new areas. These include big data and its applications to computer vision and image processing (particularly in biology and medicine), information flow in social networks, and online/mobile data protection and privacy preservation.
7 years of fruitful collaboration leads to major scientific impact at the top of the innovation chain
In 2006, Microsoft Research and Inria solidified a shared vision by opening a joint computer science laboratory with the ambition to create a world-leading public/private research centre. The new centre was innovative in both research agenda and operational model, with all results, ideas, technologies, publications, and software, being released to the international scientific community and made publicly available.
After 7 years of scientific activity, the excellence of the centre’s work and its international impact have been well-established.
- One of its main successes stems from the team headed by Georges Gonthier (Microsoft Research) who, in September 2012, completed a computer-assisted proof of the Feit-Thompson theorem, considered a “monument” of twentieth century mathematics. This acclaimed success in formal proof techniques could profoundly affect both computer science and mathematics.
- Another example is the contribution to the creation of a digital archeology start-up in 2013. The research project led by Professor Jean Ponce (Ecole Normale Supérieure de Paris and Inria) consisted of applying innovative techniques in image processing to virtually rebuild in 3D, and from simple photographs, the lost setting of one of the biggest Patrician villas of Pompeii, the Villa of Diomedes. Initially hosted within the public Parisian Incubator Agoranov, the start-up Iconem (www.iconem.com) has now flown the nest with the ambition to provide these technologies to safeguard major archeological and historical sites worldwide.
- And very recently, we have been delighted to learn that Leslie Lamport, from Microsoft Research Silicon Valley who has been contributing to the Joint Centre since its very beginning, received the 2013 Turing Award, seen by many as the “Nobel prize of Computer Science”. Such a reward not only highlights an exceptional career, but as well how collaboration is key to advancing research.
According to Michel Cosnard, CEO of Inria,
This 7 year collaboration enabled our researchers to meet major scientific challenges in a world becoming more and more digital. The partnership renewal with new research areas such as security and data protection or Big Data path the way to fruitful perspectives between our two institutions.
A commitment to four research areas
Between 2006 and 2013, digital deeply rooted itself at a fast pace in all aspects of economy, science and our everyday lives. Consequently, we have decided to extend our research domains to four new areas. We are reasserting our commitment to formal methods and are introducing the new areas of machine learning and big data, computer vision and medical imaging, and social networks and privacy.
- “How can bugs be anticipated in an all-digital world?”
Since its creation, the Microsoft Research-Inria Joint Centre has focused on applying formal method techniques to guarantee via mathematical logic that software is “correct”, i.e. free from bugs. Thanks to this approach a research team led by Cédric Fournet (Microsoft) and Khartik Barghavan (Inria) delivered a certified implementation of HTTPS, the protocol that guarantees the confidentiality of transactions on the Web. With the same tools, they discovered flaws in the standard deployments of the protocol and in the protocol itself, revealing attack mechanisms that exploit these flaws.
According to Laurent Massoulié, Director of the Microsoft Research-Inria Joint Centre: “Innovation often springs from unexpected opportunities: as such, our fundamental work on formal methods paved the way for our researchers to reinforce the security of Internet protocols such as HTTPS, a development that we had not anticipated at the onset of the project.”
Machine learning and Big Data
- “How to give meaning in a deluge of data?”
Given the scale of data that needs to be processed and the number of parameters to be taken into account, big data challenges the state of the art in machine learning techniques. In order to get a statistical model of acceptable quality, large amounts of data must be processed, such as DNA sequences for example. In this case, calculation times can become crippling. As such, the ambition of the new project supervised by Francis Bach (Inria) is to find new algorithms remaining efficient in a big data context. The aim is to reach the ideal compromise between quality of the learning and computation time.
Computer vision and medical imaging
- “How can medical treatments be tailored to the characteristics of each patient?”
Knowing how to “read” a MRI image to detect, in a timely manner, a tumor or an anomaly, or how to model an object, a place or an organ in 3D: all this is possible today thanks to analysis and synthesis of images or videos.
In their project, Nicholas Ayache (Inria), Antonio Criminisi (Microsoft Research), and their colleagues apply machine learning techniques to medical imaging. Their focus is on MRI images of the heart and brain. Using computer vision methods similar to those deployed in the Xbox Kinect device, the team intends to automatically generate annotated synthetic images, i.e. images that are realistic but do not belong to any actual patient. The potential range of applications is extremely large: from diagnostics to specific treatment, to prevention and the quantification of the effects of a drug. This makes an additional step towards personalized and preventive medicine.
Social networks and Privacy
- “How can online privacy be enhanced?”
In this new project, Laurent Massoulié and his team focus on how users can get the best access to specific information via online social networks such as Twitter and Facebook. The scientific challenge of the research lies in guaranteeing access to information which is both more relevant and more personalized whereas aggregating, sorting and filtering the data is carried out by the network. With this in mind, they will create and develop algorithms that will automatically recommend to users potential contacts with similar interests, thus enabling new connections between users. The team will also seek to create reputation mechanisms for users to facilitate the evaluation of their level of expertise and to encourage them to filter the information which they access in the best possible way.
While consumers and citizens are increasingly aware that the uncontrolled flow of data generated by the proliferation of screens in our daily lives can threaten their privacy, they also know that it can be useful to share part of their personal data in certain situations, whether individually or as a group. This can be the case for the aggregation of individual GPS positions that enable getting a precise and reliable knowledge of the traffic conditions in urban areas in real time.
In a new project, Catuscia Palamidessi (Inria) and her team will create partial disclosure models in which private information is protected and delivered either encrypted or deliberately distorted with the addition of “noise”.
These articles could interest you:
To find out more
- 100 researchers contributing to the work of the laboratory, including 30 young researchers
- 23 nationalities
- 12 public software and library releases
- 68 young researchers have been trained at the centre between 2007 and 2013
- 23 PhD theses have been defended since 2007 and an additional 18 will be defended by 2017