The goal of project DEREC is to demonstrate the feasibility of relativistic cryptography, a new and exciting research direction which takes advantage of the No Superluminal Signaling (NSS) principle in order to perform various cryptographic tasks. NSS states that no information carrier can travel at a speed greater than the speed of light. By enforcing timing and location constraints, it becomes possible to exploit the NSS principle in order to achieve information theoretic security instead of computational security for a broad class of cryptographic tasks. The advantage of relativistic cryptography is that it offers long-term security, independent of any future advance in computer software or hardware (including quantum computing). In addition to providing trust and longer lifetime of our cryptosystems, it also ensures retroactive security: a malevolent organization could store encrypted data right now and wait until it reaches sufficient computational power to decrypt them in the future. Since more and more sensitive data gets moved to the cloud, this issue is certainly of the utter importance for all actors of our modern society. Most cryptographic proposals rely on computational assumptions and therefore fail to offer long-term and retroactive security. Without being too pessimistic, we can assert that several such schemes will be obsolete in the 10 to 30 years to come. Proposals for long-term security already exist, but involve quantum hardware and therefore remain expensive and difficult to develop at large scale, despite real technological advances in the past few years. DEREC is a unique proposal that combines the best of both worlds. On the one hand, it provides information-theoretic security and therefore long-term security based on the NSS principle. Since falsifying this principle would imply the possibility of traveling back in time, we feel confident that it is as good as a cryptographic assumption can get! On the other hand, the necessary hardware to implement relativist cryptography is very standard and available today. Our cryptographic applications will use specific timing and location constraints but they can be achieved very easily in practice. In particular, we will need fast and low latency communications, but not as efficient as the already existing solutions that have been developed for high frequency trading for instance. This means that the required technology is already available and can be transferred to relativistic cryptography easily. We will focus on secure multiparty computation, with applications such as voting schemes, secret auctions, or password-based authentication schemes. Our final goal is to implement these protocols, with proven security, at the end of the 4-year project so that they can be deployed on a larger scale in the 5 to 10 years to come. We limit ourselves to a few tasks in this project for pragmatic reasons but the potential benefits of relativistic cryptography go far beyond them. Now is a perfect for the development of this technology: on top of hardware accessibility, recent preliminary results, notably by the project leader, showed that relativistic cryptography primitives can be efficiently implemented. It is a very positive and exciting time for relativistic cryptography. The team is young, dynamic and covers all the needs for the success of the project. The finality of the project will foster strong collaboration and open broad perspectives with both the academic and the industrial world. For the project coordinator, this project would be a decisive step in his career, giving him the opportunity to manage a team including students. It will also be an important step towards applying to an ERC grant. The demanded help is 225k€ + administrative overhead and will be used for funding a PhD student, a 1 year postdoc, equipment and academic travels.

The considerable diversity of long-lived magnetic fields observed in the Universe raises fundamental questions regarding their origin. Although it is now widely accepted that such fields are sustained by a dynamo instability in the electrically conducting fluid layers of astrophysical bodies, in most cases the very nature of the flow motions powering the dynamo is essentially unknown. To exhibit a flow capable of amplifying and maintaining a magnetic field is a challenging task: indeed several anti-dynamo theorems forbid the emergence of a dynamo instability out of an infinitesimal magnetic seed field in « too simple » flows (in the sense that they present too many symmetries, either for the fluid motions to sustain a dynamo, or for a magnetic field to be sustained by dynamo action). However, these anti-dynamo theorems do not necessarily extend to the stability of a magnetohydrodynamic (MHD) flow with respect to finite-amplitude dynamo seeds, which on the other hand can now be investigated with mathematical tools stemming from nonlinear variational optimisation. The Direct-Adjoint Looping (DAL) method relies on successive forward and backward time-integration of the system governing equations and their adjoint to optimise fully nonlinear, time-dependent problems. This optimal control method has proved in the last few years to be a powerful ingredient to study the subcritical transition to hydrodynamic turbulence. The present research plan aims at developing the numerical tools required to apply the nonlinear DAL method for the first time to full, unsteady MHD flows. This approach will allow to study minimal dynamo seeds (or in other words, the amplitude and spatial structure of the smallest magnetic legacies that can trigger a subcritical dynamo) in simple swirling flows relevant to stellar systems. Furthermore, nonlinear optimal control will be used as a physical diagnostic to gain novel understanding of the mechanisms that are most favorable to self-sustained dynamo action in astrophysical flows.

The GETRF (Gestion Efficace des Transmissions dans les Réseaux sans Fil) project proposes to study new mechanisms to improve the performance of wireless multihop military networks. This project considers wireless multihop networks as a whole of which mobile ad hoc networks (MANETs) and of sensor wireless networks (WSNs) are specific cases, with both differences and commonalities. The goal is to address two major challenges in tactical military networks: 1. efficient management of energy, latency and network capacity. 2. efficient management of the network in mobile or degraded environment. These objectives may for instance answer two scenario of functioning of military tactical networks: on one hand, a stabilized, static, situation in operating normally, where the goal is to optimize the network usage ; and on the other hand, the case of more critical scenarios, for instance, the case of high mobility, the case of unexpected degradation of the quality of one part of the network, and, also, the case where the broadcast information is exceeding the capacity, which are all cases where the goal is to operate despite the very high constraints. To answer the objectives, three type of mechanisms will be studied in the project: - network coding techniques. These recent schemes allow for more robust information broadcast in the case of changes in the network (mobility, degradation of one part of the network), to reduce the control traffic exchanges, and more generally, to increase network efficiency (less transmissions, more capacity). - coloring techniques. They allow time-slots to be assigned to nodes or links of the wireless network. These techniques allow one to build efficient Time Division Multiple Access techniques. The medium access scheme obtained can be very energy efficient and can also offer very good end-to-end delays - the opportunistic routing schemes. Opportunistic routing protocols, usually based on geographic routing, can improve the total throughput of ad hoc networks. They can also be combined with medium access schemes to build network nodes operating in low-duty cycle mode. These schemes can be very energy efficient. Another application of opportunistic routing is a coupling with mobility (mobile opportunistic routing schemes). On can take advantage of nodes' mobility to optimize the delivery delay.

In the near future, intelligent and autonomous systems will become more ubiquitous and pervasive in applications such as autonomous robotics, design of intelligent personal assistants, and management of energy smart grids. Although very diverse, these applications call for the development of decision-making systems able to interact and manage open-ended, uncertain, and partially known environments. This will require increasing the autonomy of ICT systems, which will have to continuously learn from data, improve their performance over time, and quickly adapt to changes. EXTRA-LEARN is directly motivated by the evidence that one of the key features that allows humans to accomplish complicated tasks is their ability of building knowledge from past experience and transfer it while learning new tasks. We believe that integrating transfer of learning in machine learning algorithms will dramatically improve their performance and enable them to solve complex tasks. We identify in the reinforcement learning (RL) framework the most suitable candidate for this integration. RL formalizes the problem of learning an optimal control policy from the experience directly collected from an unknown environment. Nonetheless, practical limitations of current algorithms encouraged research to focus on how to integrate prior knowledge into the learning process. Although this improves the performance of RL algorithms, it dramatically reduces their autonomy. In this project we pursue a paradigm shift from designing RL algorithms incorporating prior knowledge to methods able to incrementally discover, construct, and transfer “prior” knowledge in a fully automatic way. More in detail, three main elements of RL algorithms would significantly benefit from transfer of knowledge. (i) For every new task, RL algorithms need exploring the environment for a long time, and this corresponds to slow learning processes for large environments. Transfer learning would enable RL algorithms to dramatically reduce the exploration of each new task by exploiting its resemblance with tasks solved in the past. (ii) RL algorithms evaluate the quality of a policy by computing its state-value function. Whenever the number of states is too large, approximation is needed. Since approximation may cause instability, designing suitable approximation schemes is particularly critical. While this is currently done by a domain expert, we propose to perform this step automatically by constructing features that incrementally adapt to the tasks encountered over time. This would significantly reduce human supervision and increase the accuracy and stability of RL algorithms across different tasks. (iii) In order to deal with complex environments, hierarchical RL solutions have been proposed, where state representations and policies are organized over a hierarchy of subtasks. This requires a careful definition of the hierarchy, which, if not properly constructed, may lead to very poor learning performance. The ambitious goal of transfer learning is to automatically construct a hierarchy of skills, which can be effectively reused over a wide range of similar tasks. Providing transfer solutions for each of these elements sets our objectives and defines the research lines of EXTRA-LEARN. The major short-term impact of the project will be a significant advancement of the state-of-the-art in transfer and RL, with the development of a novel generation of transfer RL algorithms, whose improved performance will be evaluated in a number of test beds and validated by a rigorous theoretical analysis. In the long term, we envision decision-making support systems where transfer learning takes advantage of the massive amount of data available from many different tasks (e.g., users) to construct high-level knowledge that allows sophisticated reasoning and learning in complex domains, with a dramatic impact on a wide range of domains, from robotics to healthcare, from energy to transportation.

Cryptographic protocols are an essential building block to secure online communications. Relying on cryptographic primitives, such as encryption and digital signatures, these protocols typically guarantee security properties such as confidentiality and authenticity of communications. These properties are for instance basic goals of TLS, the most widely deployed cryptographic protocol, underlying all https Internet connections. Many modern applications may also need to guarantee properties related to users' privacy: such properties include anonymity and unlinkability (users cannot be traced) properties. History has however shown that cryptographic protocols are very error-prone and attackers have been able to exploit design and implementation flaws. The difficulty of designing and deploying secure protocols comes from an inherent asymmetry in security: while a protocol designer needs to think about all possible attacks, an attacker only needs to find a single weakness. Even on rather simple protocols, it is difficult for a human to explore all possible cases in a security proof when we take into account an attacker that has control of the underlying communication network and can intercept, modify and insert messages, in addition to the concurrent nature of the protocols. This task is even more complicated when requiring strong security properties that hold, or guaranteeing at least a degraded form of security, when parts of the system under study have been compromised. Therefore, it is essential to design algorithms that are able to automate security proofs, or automatically detect attacks in protocols. This approach, where messages are represented as first-order terms, in a symbolic model, was pioneered by Dolev and Yao and has been extremely successful, e.g., when analyzing new, upcoming security standards such as TLS 1.3 or 5G. The goal of this project is the development of efficient algorithms and tools for automated verification of cryptographic protocols, that are able to comprehensively analyse detailed models of real-world protocols building on techniques from automated reasoning. Automated reasoning is the subfield of AI whose goal is the design of algorithms that enable computers to reason automatically, and these techniques underlie almost all modern verification tools. Current analysis tools for cryptographic protocols do however not scale well, or require to (over)simplify models, when applied on real-world, deployed cryptographic protocols. We aim at overcoming these limitations: we therefore design new, dedicated algorithms, include these algorithms in verification tools, and use the resulting tools for the security analyses of real-world cryptographic protocols. The resulting tools will have increased efficiency, better automation and a wider scope. These developments will be driven by and validated on three classes of protocols: e-voting protocols, protocols for mobile devices from the 5G standard, and protocols for modern messaging applications, in particular the NOISE protocol framework. Our research is driven by the development of verification tools and their application to real-life protocols. Developing efficient, usable tools requires to develop solid, theoretical foundations, on the one hand, and evaluation on real-life case studies, on the other hand, to understand their limitations in practice. Our project will therefore be structured around the development of several, complementary tools that are developed in the PESTO research team, headed by Steve Kremer. On a more technical side, the methods developed in the project include symbolical reasoning on partially ordered traces, new algorithms for efficient subsumption in first-order Horn clause resolution, the design of equational reasoning techniques to enable support of associative, commutative operators and heuristics for guiding the state exploration to favor termination.