Loreley: towards decentralised and secure collaborative systems

Collaborative systems struggling in the face of scale change

“The first collaborative systems were designed in the 1990s to allow a few dozen users to interact on fairly simple tasks”, explains Claudia-Lavinia Ignat, Head of the Loreley project team. “But today we’re looking at a different scale. Some open-source projects involve very large numbers of developers, while Wikipedia pages can be modified by a multitude of editors at the same time during major events”.

In such cases, current tools are reaching their limits in terms of both the number of simultaneous users and security. Current solutions rely on a central server that receives, decrypts and stores every user’s data, which creates a weakness in the system. “If this server is attacked, all the data is compromised”, explains Claudia-Lavinia Ignat. “What's more, this kind of approach also raises data confidentiality and sovereignty issues, since data is generally entrusted to a single large company”.

Towards peer-to-peer collaboration

Loreley will combine expertise in computer science, distributed systems and CSCW (Computer-Supported Cooperative Work), a field of research that takes the human factor into account to propose solutions tailored to users. By bringing together these different fields, they aim to achieve a central objective: develop large-scale collaborative peer-to-peer systems (without a central authority) that are distributed, resilient, secure and built on trust between the stakeholders. This would mean that data control is in the users’ hands, thus offering a number of advantages: firstly, without the need for a central server, deployment costs are lower; secondly, users can share data only with those they trust; thirdly, the fragmented and encrypted data is stored on each user’s computer, which reduces the risk of breach and failure (if a node is attacked or becomes deficient, only the data on that node is compromised). 

Meeting the challenges of secure distributed computing that is adapted to people’s needs

However, there are still many technical hurdles to overcome, starting with the development of algorithms capable of merging data from different locations and taking user needs into account. “For example, we can't simply reject one user's modifications if they are not compatible with those of another”, explains Claudia-Lavinia Ignat. “We need to keep them so that they can be reworked”.

In terms of security, the algorithms need to be adapted here too. “There are many encryption solutions already out there, but they don't consider mutable data, i.e. data that can be changed by users", the Loreley manager continues. “To prevent intrusions, we also have to design a system that tells each user who has the right to be part of the group, without this information passing through a central authority”.

To add an extra difficulty, Loreley hopes to get a group of humans and computers working together in a system of this kind, where the former edit a shared document, for example, while the latter carry out spell checking, translation or summarising. “The issue of trust between humans but also between humans and computers is crucial to effective interaction”, says the researcher. “The human factor is a common theme in all our research. We have to take it into account to ensure several objectives are achieved, namely that the tools are appropriately designed and adapted, that both trust and conflicts can be managed, and that the quality of use can be evaluated, in addition to integrating the group effect on each user’s behaviour”.

From peer-to-peer networks to distributed AI

To meet these challenges, Loreley is counting on a number of corporate partnerships, such as the Inria Alvearium challenge, which brings together five Inria teams with the start-up Hivenet and aims to develop a peer-to-peer cloud. “Hivenet proposes to exploit the unused capacity of each user's equipment to carry out complex tasks in exchange for monetary compensation”, says Claudia-Lavinia Ignat. “The company already has a large number of clients”. What does the project involve? “At the moment, the cloud only stores fixed data, so we are working on mutable-data management and group key management mechanisms so that users can be added and removed. We’re also studying peer-to-peer network attacks and have uncovered a flaw which we are hoping to find a solution for”.

Along a similar line, Inria and Hivenet have also launched the Cupseli challenge, which involves 11 Inria teams, including Loreley, and aims to use peer-to-peer tools to allow AI to be used without a centralised server in order to save energy, in particular. “There are still a lot of technological stumbling blocks, which is why we joined forces with Inria: their teams’ incredible expertise enables us to build robust and reliable technological foundations together", says Alexandru Dobrila, a research engineer at Hivenet. “These projects allow us to compare our work with best practice in each field, make adjustments or sometimes even start again from scratch based on research results”.

The need for collaborative systems in all sectors

Loreley is also working with other French companies to pursue work on Mute, a collaborative peer-to-peer editor developed by the Coast Inria team, which enables end-to-end data encryption. 

It is also involved in the European DXP (Data Exchange Platform) project, which is run in collaboration with Amadeus, a travel services management and distribution company. “Here, we’re using edge computing to design a federated and distributed data exchange platform that allows data to be processed and stored as close to its source as possible”, says Claudia-Lavinia Ignat. 

Lastly, the researcher also coordinates a project within PEPR eNSEMBLE called Pilot: “The idea is to revisit the sociotechnical environment to identify the needs of different sectors - health, industry, education, crisis management, etc. - in order to create new conceptual models for long-term collaboration. In the case of healthcare and patient records, for example, we have to know which professionals need access to what types of data, but also when and how, etc.” 

A total of 25 teams are working on this project over eight years, until 2030, with a budget of five million Euros. The project involves a wide variety of disciplines including psychology, sociology, management and computing, and is a perfect illustration of Loreley's objective: putting people at the heart of distributed and secure collaborative systems. 

(*) The project team Loreley is a joint venture between the CNRS, Inria and the University of Lorraine based at the University of Lorraine Inria Centre and the Lorraine Laboratory for Research in Computer Science and its Applications (CNRS/University of Lorraine)