Exploratory research

Partner laboratory: LGI – CentraleSupélec

Until now systems have been designed to address users’ needs. For example, cars are currently controlled by humans. However, nowadays, needs tend to evolve towards cooperation between the system and humans in order to achieve a common goal. In this perspective, it is fundamental that a system can jointly evolve with humans over time by taking into account human factors. It then becomes essential to integrate human factors aspects and their temporal evolution all along the system’s design process. Humans are an integral part of the system. They are a subsystem in the global System of Systems. Researchers from the Human System Integration (HSI) domain proposed extensions to the SysML language, allowing human component to be integrated all along the system’s design phases. Nonetheless, the question of the human’s temporal dynamics and the integration of human factors in the choice of an appropriate architecture are two axes of research that remain in discussion. This exploratory research will mainly focus on these last two points.

Partner laboratory: L2S – CentraleSupélec

Reconfigurable Intelligent Surfaces are artificial materials that allow to control, using software, the way in which impinging electromagnetic waves are reflected. The propagation environment becomes hence a Smart Radio Environment, capable to fashion itself to favor the communication task, enabling transmission with very small amounts of energy.

Partner laboratory: Ulm University

The objective of this exploratory research project is to carry out a taxonomy of the existing privacy protection mechanisms in the Vehicular Edge Computing (VEC). This will be ensured by a fine understanding of the context, the algorithms and the objectives of these mechanisms, while identifying their strengths and limits.

Partner laboratories: L2S – CentraleSupélec, School of Computer Science and Engineering – NTU Singapore

The focus of this exploratory research consists in addressing the resource allocation problem in the Internet of Things (IoT) domain. Indeed, we will address the placement of chains of Virtual Network Functions (VNF) for IoT applications to jointly optimize the network resources and the computational resources (processing load) of connected nodes in the edge. The main challenges that have the merit to be addressed in this proposal will cover the IoT VNF chains placement when considering limited network resources and processing operation to be performed jointly. This problem is NP-Hard, and then necessitates to investigate combinatorial optimization techniques and Game Theory approaches to converge towards near optimal solutions.

The proposed approaches will be implemented on a testbed to show their efficiency and to benchmark them with other existing approaches.

Partner laboratories: LIP6 – Sorbonne Université, Dept. of Statistics – Stanford University

The discovery of hyperbolic geometry was a fairly important event in the history of mathematics and science in general. Indeed, hyperbolic spaces have many applications in various branches of science such as general relativity (with Minkowski's notion of space-time), in cosmology (FLRW models are the main candidates for modeling the shape of the universe as a whole), in number theory (modular shapes have solved several number theory riddles) etc. In recent years new applications of hyperbolic geometry in the field of artificial intelligence and deep learning have emerged.

Several research studies have shown that hyperbolic spaces are more capable of capturing complex data such as graphs, texts and images than their Euclidean counterparts.

In this context, the project "Machine learning with hyperbolic neural networks" aims to

• explore the efficiency of hyperbolic neural networks for concrete applications,

• compare their performances with Euclidean neural networks,

• determine whether hyperbolic algorithms provide more robust, interpretable and safe results than their Euclidean analogs.

This project is carried out in collaboration with Université de Sorbonne and Stanford university.

Partner laboratory: Equipe Kopernic – Inria Paris

This project is focusing on studying the communication protocols in unmanned aerial vehicle (drones) and it tries to propose a protocol that takes in account execution deadlines as they are used in the real-time embedded systems. The idea of this research topic came from the fact that a large number of drone applications (e.g. autopilot) are open source and the steps of their conception can be tracked for a better understanding of its functioning.

Through this research subject, IRT SystemX will be able to consolidate its relations with the real-time community and explore the possibility of further project proposition on the subject of real-time embedded systems and cyber-physical systems. For this work, we collaborate with the Kopernicus team at Inria Paris which has as its main fields of study the real-time embedded systems and probabilistic timing analysis.

Partner laboratories: Samovar – Télécom SudParis, Equipe pir2 – Inria Paris, Dept. Computer Science – Université du Luxembourg, Dept. of Electrical and Electronic Engineering – City University of London

Public-key cryptography is one of the most important foundations of modern cyber security. The security of the Transport Layer Security (TLS) protocol that protects billions of Internet connections daily relies on public-key encryption and digital signatures. This protocol protects the confidentiality and integrity of billions of Internet connections daily, for example in e-banking and e-commerce applications, and email traffic. However, as a result of Peter Shor shows, the public-key schemes that are being used today will become insecure once quantum computers reach maturity.

This aim of this project is to study and develop Post-Quantum Cryptographic (PQC) primitives of key-exchange, encryption and digital signature that can replace their classical counterparts due to being able to render the attacks of quantum adversaries ineffective. The core idea is to design cryptographic primitives whose security rely on problems that cannot be solved by quantum computers in admissible time. Additionally, we are interested in machine-checked proofs of security for a variety of classical cryptographic primitives to the post-quantum world. This project also proposes post-quantum scheme taking in account their possible application to Blockchain security.

Partner laboratory: LAAS – CNRS

In the automotive world, dependability guarantees the functional safety of electronic / electrical systems in vehicles. This ensures that the services that operate the vehicle should not cause harmful effects. However, there is another concept which is essential: cybersecurity. In fact, it represents all of the mechanisms that protect the vehicle from external and intentional threats. There is a strong link between these two notions. Indeed, their coordination would make the vehicle more resilient. The objective would then be to define the possible links between safety and cybersecurity so as to determine the essential elements for the validation of Intelligent Transport Systems, in particular autonomous vehicles.

To do so, we will define a state of the art concerning safety / cybersecurity hybridization, in order to propose a scientific article of the "survey" type. Then, we will work on different case studies around IRT SystemX platforms (MOSAR, automotive platform of the CTI project).

The ultimate aim of this Exploratory Research project is notably to be a good starting point for the construction of new projects at IRT SystemX revolving around this subject matter and to use its expertise to propose its own approach.

Partner laboratory: Samovar – Télécom SudParis

In recent years, machine learning algorithms have massively integrated the defense arsenal made available to professionals in the security sector. A perfect illustration of this is the detection of intrusion. Although the results and performance of Machine Learning are constantly improving, it is still relatively easy to deliberately mislead it, by means of what is commonly known as Adversarial Machine Learning Attacks. The proliferation of adversary attacks raises the question of "who will really benefit from machine learning in the future?" Does it benefit attackers or security professionals?

The objective of this exploratory project is to study the different classes of machine learning algorithms used in intrusion detection, demonstrate the impact of adversary attacks on the performance of intrusion detection systems and propose solutions that would make learning models more robust vis-à-vis adversary attacks and in order to make more reliable predictions.