Scientific Themes and Challenges

Cyber-security

The cyber-security of systems and networks is now a fundamental challenge. The RISC department addresses this theme through multiple scientific axes and across various application domains. One of the department’s strengths is its ability to approach information system security from three complementary perspectives: software security, hardware security, and network security. The proposed security solutions can thus address both software (applications or operating systems, such as in critical embedded systems or IoT devices) and hardware (e.g., cryptographic mechanisms integrated directly into architectures supporting in-memory computing). This cross-cutting vision of information system security is a significant asset, given the strong interdependence between software and hardware in modern computing systems.

In the field of network security, the department explores multiple complementary aspects. It has developed strong expertise in intrusion detection using unsupervised learning models, a focus maintained for many years, and more recently in the generation of synthetic training data. These efforts are conducted in collaboration with industry partners, enabling the integration of the laboratory’s scientific results into commercial security products. Other notable work focuses on the security of IoT communication protocols and their lower layers, addressing both offensive and defensive approaches. These contributions have been published in major security conferences in recent years.

Virtualization and Softwarization

In the era of resource virtualization—extending from cloud servers to the edge of communication infrastructures—the range of possibilities has expanded significantly. However, these advancements come with unprecedented complexity, making the design, planning, deployment, and monitoring of communication networks and their applications increasingly challenging. The boundary between cloud application resources and the communication networks that interconnect them is becoming increasingly blurred. Our research revolves around two complementary themes:

First, we study the Cloud Continuum, which encompasses all interconnected cloud environments, including edge infrastructure and hybrid systems, particularly those related to the Internet of Things. Our contributions have explored new approaches for monitoring applications, data centers, and Internet exchange points, aiming to enhance the autonomy of communication systems.

Second, our research focuses on the orchestration of application and network services to optimize their delivery. We are particularly interested in the dynamic description, discovery, composition, and configuration of services, leveraging IoT service virtualization and the optimal placement of virtualized network functions (VNFs) and their slices, up to their deployment. This work also extends to multi-domain environments, where interconnection and orchestration between heterogeneous infrastructures present new challenges in terms of consistency, compatibility, performance, and resource management.

Toward Trustworthy and Frugal Artificial Intelligence

Over the past decade, methods grouped under the umbrella of artificial intelligence have achieved spectacular results in numerous fields. However, deploying these algorithms requires guaranteeing properties that go far beyond traditional performance criteria, particularly to ensure their safety and reduce their ecological impact.

The RISC department emphasizes trust and frugality in AI algorithms. Trust is addressed through two complementary aspects: transparency in AI models, essential for understanding and justifying their decisions, and runtime monitoring of AI model behavior. Frugality, on the other hand, aims to design algorithms that consume fewer data, computational resources, or energy.

While AI provides tools to address some of the problems studied by the department, we also highlight an inverse movement: traditional departmental tools are applied to improve AI, both at its core and in its implementation and integration as a system. RISC researchers have combined their AI expertise with experience in dependability, computer networks, and stochastic modeling.

Analysis and Engineering of Critical Systems

The concepts of analysis and systems are central to the research of all LAAS departments, as reflected in the laboratory’s acronym. Within the RISC department, we address these issues by focusing on critical systems, whose failures or violations of operational constraints can have severe consequences. We also consider how these systems are designed, validated, and scaled, hence the use of the term engineering. In this context, we examine systems not only during their operation but throughout their lifecycle, from design and development (with an emphasis on modeling, testing, simulation, and formal verification) to long-term monitoring of their performance and reliability.

Our work combines fundamental approaches with practical studies, aiming to provide innovative and robust solutions that enhance the reliability, predictability, and performance of systems, particularly in aerospace, transportation, energy, and critical infrastructure sectors.

Research in this area covers several key axes. We develop advanced methods for software testing, focusing on early defect detection and safety property validation. We also use safety arguments and formal verification to assess and quantify confidence in systems. Our research includes the study and scaling of real-time systems to meet the performance and responsiveness requirements of critical applications. Finally, we address these issues from the perspective of resilience and quality of service (QoS) compliance, as demonstrated by our work on proactive and adaptive strategies for managing QoS in the cloud continuum to optimize the performance of distributed systems.

Distributed Algorithms

Research in this area focuses on the design and analysis of distributed algorithms in computer networks and distributed systems. The scope is broad, ranging from theoretical studies of the properties of certain distributed algorithms (e.g., consensus) based on communication graphs to more applied research on parallel algorithms, federated learning, and distributed resource allocation in cellular networks, data center networks, and LPWANs. Major results have been achieved, particularly in fast local rerouting in networks and in measuring and improving the performance and robustness of Internet routes.

Methodologically, our research is based on a wide range of theoretical tools, including graph theory and mathematical optimization, as well as more original approaches such as combinatorial topology, stochastic optimal control, probabilistic analysis, and game theory. The emphasis is on the theoretical analysis of algorithm properties and the derivation of worst-case performance guarantees. However, experimental validation of the proposed algorithms is not neglected; our work relies on either in-vitro data from experimental platforms or in-vivo measurements from operational networks and systems.