On March 15, 2022, 2-4 pm, we will host three seminars on fronteers in IoT QoS and constrained computing management.
Room: 33.1.19, 2 rue conté, Paris.
Speaker: Nabil Makarem, AUB, Lebanon
Title: Towards an Efficient Congestion Control for the Constrained Application Protocol in IoT networks
Abstract: IoT has recently attracted attention in almost all areas such as agriculture, industry, homes, health care ,and transportation. In this context, healthcare IOT includes tracking and monitoring patients, chronic diseases and even remote services. In this kind of applications, usually researchers tend to reduce losses and improve packet delay. Another important application where our research can apply is disaster management. Disasters have destructive impact on economics and human life. However, IoT which has proven to be capable of providing solutions for disaster management. Here also, performance is critical in terms of packet loss and latency. In this talk, I will talk about the importance of improving performance in IoT networks, the challenges in such environments and the reasons behind focusing on CoAP which is one of the main candidates for a lightweight communication protocol for the Internet-of-Things. The simple congestion control mechanism in this protocol significantly reduces its performance especially in networks with high packet loss, and thus preventing an efficient deployment of the protocol. So, secondly, I will talk about improving the retransmission timeout estimation for congestion detection, and adopting adequately a rate-based approach for congestion counteraction, while maintaining simplicity required by constrained devices.
Bio: Nabil Makarem received his Master’s degree from the Lebanese American University in 2014 and his PhD degree in computer science from Sorbonne University and the Lebanese University in 2021. His current research area is the Internet of Things, with an emphasis on performance evaluation and improving congestion control mechanisms in IoT Networks. He has worked in several universities and corporations, holding different positions such as System and Network Engineer, IT Manager. Since 2019, he is an instructor/lecturer at the American University of Beirut.
Speaker: Lemia Louail, Université Ferhat Abbas Sétif 1
Title: QoS optimization in IoT systems
Abstract: Internet of Things refers to the interconnection of everyday objects through Internet. These objects (sensors, cars, refrigerators, …) are often equipped with ubiquitous intelligence and communicate with each other in order to improve the quality of our lives.
Several domains are integrating the IoT concept such as wireless sensor networks, Industries, etc.
Quality of service (QoS) refers to metrics which describe the overall performance of a system such as latency, energy consumption, packet loss, availability and many other metrics. QoS is particularly important because it indicates whether the system is well behaving or not.
This presentation introduces some examples of QoS optimization in a number of IoT systems such as: Latency optimization in WSNs using cross-layer approaches, Sink placement in WSNs, Task scheduling in Edge/Fog/Cloud architecture, Latency and energy optimization in nano-networks,
Bio: Lemia Louail received her PhD in 2016 from University of Franche-Comté Besançon in France. She is currently an associate professor at University of Sétif 1 in Algeria. Her main research interests include networks and distributed systems, communication protocols and IoT systems.
Speaker: Chafiq Titouna, University of Paris-Cité
Title: Securing Unmanned Aerial Systems against cyber-attacks
Abstract: With the growing use of Unmanned Aerial Vehicles (UAVs) in military and civilian applications, cyber-attacks are increasing significantly. GPS spoofing and False Data Injection (FDI) attacks are considered the most common attacks against Unmanned Aerial Systems (UAS). Therefore, in order to ensure the predefined mission of UAVs, the detection of these attacks has become a real challenge that allows achieving a high level of flight security, high reliability, and schedule maintenance in time. In this talk, we will present two techniques for the detection of the previous attacks. Firstly, we will describe the UAS and some types of threats/vulnerabilities. Then, we will present in detail our proposals including the preliminary simulation results. Finally, we will conclude this talk with some future works.
Bio: Chafiq Titouna is a temporary research assistant at the University of Paris. He received his Ph.D. degree in computer science in 2017 from the University of Bejaia in collaboration with the University of Versailles, a Magister degree in computer science in 2012 from the Higher National School of Computer Science, and a State Engineering degree in computer science from the University of Batna in 2008. His current research interests include anomaly detection, malware detection and cyber-security in WSN, IoT and UAV.
Speaker: Makhlouf Hadji, SystemX
Title: Virtualized Face Detection at the Edge – Reinforcement Learning based optimization.
Abstract: Real-time requirements in video streaming and processing are increasing and represent one of the major issues in industry 4.0 domains. In particular, Face Detection (FD) use-case has attracted the interest of industrial and academia researchers for various applications such as cyber-physical security, fault detection, predictive maintenance, etc. To ensure applications with real time performance, Edge Computing is a good approach which consists in bringing resources and intelligence closer to connected devices and hence, it can be used to cope with strong latency and throughput expectations.
In this presentation, we consider optimal routing, placement and scaling of virtualized face detection services at the edge. We propose an edge networking approach based on Integer Linear formulation to cope with small problem instances. A reinforcement learning solution is proposed to address larger problem sizes and scalability issues. We assess the performance of our proposed approaches through simulations and show advantages of the reinforcement learning approach to converge towards near-optimal solutions in negligible time.
Bio: Makhlouf HADJI received his PhD degree in computer science and applied mathematics from Télécom SudParis, jointly with University of Paris. He was a research fellow at Télécom SudParis for 3 years. Next, he joined IRT SystemX as senior researcher. In March 2017, he obtained his docent habilitation (HDR) at Paris University. Then, he joined the Scientific Department of SystemX as the Head of « Digital Infrastructure and IoT » research team. His main research interests include network virtualization and optimization, data driven algorithms, resource allocation, game theory, mathematical programming applied to device-edge-cloud continuum, IIoT and networks of future. He was involved in many European projects (ODISEA, East-Clouds, XLCloud, SAIL, …) and actually leading one Horizon Europe proposal (named : Cognitive Data Driven Orchestration of Edge-Cloud Continuum) and actively participating in two other proposals on orchestration and networking of the device-edge-cloud continuum. These proposals are prepared with industrials (ATOS, RTE, EDF, Thales, Alliander, …) and academics (TUDelft, Norwegian Center of research, TNO, Fraunhofer, …).
Speaker: Akram Hakiri, Univ. Carthage
Title: SDN-based Middleware for Control and Management of Wireless Multi-RAT
Abstract: Real-time requirements in video streaming and processing are increasing and represent one of the major issues in industry 4.0 domains. In particular, Face Detection (FD) use-case has attracted the interest of industrial and academia researchers for various applications such as cyber-physical security, fault detection, predictive maintenance, etc. To ensure applications with real time performance, Edge Computing is a good approach which consists in bringing resources and intelligence closer to connected devices and hence, it can be used to cope with strong latency and throughput expectations.
In this presentation, we consider optimal routing, placement and scaling of virtualized face detection services at the edge. We propose an edge networking approach based on Integer Linear formulation to cope with small problem instances. A reinforcement learning solution is proposed to address larger problem sizes and scalability issues. We assess the performance of our proposed approaches through simulations and show advantages of the reinforcement learning approach to converge towards near-optimal solutions in negligible time.
The 5th generation of wireless mobile network is envisioned to provide new services ranging from a small group of users, such as self-assembling robots, to mass-market services, such as mixed holographic reality, cellular-connected drones, autonomous supply chain, and massive twinning. Such services have high-performance requirements (e.g., strict requirements on ultra-low latency and reliability for mission critical communications) and make use of multiple Radio Access Technologies (multi-RATs), such as 3G, 4G, 5G, Wi-Fi, WRAN and even LoRaWAN or NB-IoT. Multi-RATs are composed of complex infrastructures with high consequences for downtime due to failures or security attacks and must deal with the added complexity of massive geographical scale. Although several efforts have focused on integrating and unifying different RATs (e.g., IEEE 802.11, 4G LTE and now 5G NR modem), however, the interoperability issues between them may slow down the pace of such deployments, network upgrades and introduction of new services. Additionally, recently the 3GPP 5G specifications supports multi-access integration in the core network with the help of a unified 5G Core. However, there is a need to extend this capability to RAN and evolve a RAN level architectural framework for unification and integration of multiple RATs, such as, IEEE 802.11, IEEE 802.22, 4G LTE and 5G-NR. Such a unified and integrated multi-RAT RANs should be able to further improve the network performance and should have the potential to simplify the interworking mechanism between RATs.
In this talk, we present a vendor independent SDN-based architectural framework that introduces an abstraction layer over the data plane, also called an SDN middleware. The SDN middleware aims to facilitate a unified control and management of data plane elements including those of IEEE 802.11 based WLANs within a multi-RAT RAN. This framework is expected to bring several advantages of SDN paradigm to multi-RAT RAN, e.g., dynamic control of access network elements through open programmable interfaces, optimized resource utilization and finer/granular control over deployment of policies/services in the access network. Additionally, we introduce our recent work on deep reinforcement learning for task offloading in SDN-enabled IoT network. Finally, we will present our ongoing work to enable Network Digital Twin (DT) for achieving resilient 5G internet services for ensuring high availability, openness, and disruption tolerance. The DT creates a virtual replica of the current network state (aka, the twin) to simplify the deployment and management of new 5G services
Bio: Dr. Akram Hakiri is an assistant professor of Computer Sciences at the University of Carthage, and a research scientist at SYSCOM Labs’ ENIT, both in Tunisia. He obtained his M.S. of computer sciences in 2008, and his Ph.D. of computer sciences in 2012, both at the University of Toulouse, France. He was also a senior scientist and researcher at LAAS-CNRS, Toulouse, France, and a visiting research scientist at the Institute for Software Integrated Systems (ISIS) at Vanderbilt University, Nashville, TN, USA. He is a visiting fellow at Santa Clara University, CA, USA, visiting senior scientist at IIT Bombay, Mumbai, India, and visiting researcher at Polytechnic University of Bucharest, Romania.
Dr. Hakiri current research focuses on developing novel solutions to emerging challenges in SDN/NFV for 5G network and beyond, network slicing and resource allocation, Digital Twins for holistic management of resilient edge-cloud 5G networks, SDN middleware for SDN based middleware for the control and management of multi-RAT technologies, cognitive edge network programmability and autonomous adaptability, Blockchain-based secure and trustworthy 5G massive IoT network and machine-learning for IoT edge resource offloading.
Dr. Hakiri is a Senior IEEE member, ONF Ambassador, ACM professional member, and the vice-chair of the IEEE COM NetSoft SDN MCM P1930.1 standardization project. He has been providing talks and presentation in different international universities in the USA, France, India, Romania, and Tunisia. He is serving as a peer reviewer for around twenty top-rated journals, technical program committee for top-venue international conferences such IEEE WCNC2021, IEEE UIC2020. Dr Hakiri has also developed an excellent publication record in top-rated journal (e.g., IEEE-Communication Magazine, Elsevier Computer Networks) and top-venue conferences (IEEE NetSoft, ACM DEBS, etc.).
