Reliability vs Security in AI-aided Smart IoT-Edge Continuum
Monday, 3 June 2024, 14:00-17:30, Room Gorilla Room 4
Speaker:
- Mauro Biagi (Univ. Rome La Sapienza, IT)
- Valeria Loscrì (Inria, FR)
Motivation and Context
Abstract – In the last years there has been an accelerated impulsion on wireless technology, making the wireless landscape richer and more complex. There are a plethora of co-existing wireless communication technologies, and the Internet of Things (IoT) is evolving and encompassing not only connected objects, but also processes, data and people. In this complex systems, it is paramount to guarantee the reliability as well as security, since our daily life is related to services and applications provided by such a kind of systems. Moreover the role played by Artificial Intelligence is upcoming an element of key discussion since it can grant more secrecy as well as it can be used for more sophisticated attacks.
Objectives – The main objectives of this tutorial consist in providing a presentation of the main wireless communication technologies and highlight their unique features and how they make the communication systems resilient and robust. On the other hand, intelligence is more and more characterising modern wireless communication architectures, with a massive use of Artificial Intelligence (AI) mechanisms as those related to Machine Learning (ML) approaches to make these systems adaptive, also in a high dynamic environment and reducing at minimum the human intervention. The main point is that intelligence can be also used by potential offender, to create new and more efficient cyber attacks or improve existing ones. in this tutorial we aim to reveal this double nature of ML approaches and we will provide some concrete examples in the IoT-Edge context.
Audience Target audience is characterized by researchers on wireless communications interested in IoT and cyber security in the context of wireless networks. Moreover people already working on IoT, Edge architectures, as well as industry people relying on Industrial Internet of Things (IIoT) based solutions can be part of the audience.
The paradigm shift related to IoT-Edge-Cloud Continuum, recently adopted and pushed by the European Commission and considered in several European Project and framework is a key concept for the evolution on 6G. Nevertheless, this paradigm makes in sort that a revision of the main communication technologies in respect of this paradigm needs to be considered. Moreover, this opens new breaches and vulnerabilities in correspondence to the low-end nodes as highlighted by ENISA [9] and NIST [10], that will be analysed in the tutorial. We believe that these concepts are at the core of EuCNC & 6G Summit.
Structure and Content
Although in the Literature IoT is not new, it is still a hot topic and, more, from the implementation point of view we are still in the dawn of IoT services and applications. An outline of the tutorial content, including its tentative schedule and the presenter for different parts is provided to the following;
- LoRa (20min) LOng-RAnge communication will be pre- sented in its essentials by focusing both on physical layer pillars (including time and frequency representation of signals) and networking one by focusing on the prob- lem, linked to security issue, of multi-user interference / multiple access control strategy [1], [2]. Real data will be shown due to an Italian company (Unidata) operator providing some key measures, and key performance in- dicators.
- SigFox (20 min) SigFox characteristics will be considered by underlining the key elements and vulnerabilities in terms of spectrum usage. The role played by cooperative detection will be also underlined together with physical layer and access strategies. The flat network architecture will be detailed by specifying pro and contras of each networking sub-system/element.
- NB-IoT (20min) NarrowBand IoT in the framework of LTE will be presented by organizing the talk into two sub-sections, one dedicated to the Downlink and one to the Uplink. Although the physical layer features may be of interest, more relevance will be given to the three operation modes (in-band, guard-band and stand-alone) as well as to channelization. In fact, despite of the other previous two IoT implementation, NB-IoT incorporates the typical features of a cellular network including power control and logical channels for handling the signaling. In this context details will be shown so as to pave the way to considerations about authentication vulnerabilities.
- Other possible enabling techniques and very specific applications (MQTT, VLC, underwater) (30min) Very specific solutions and applications may be presented by underlying why some technologies are used and what are the strength points as well as vulnerabilities. Indoor Optical wireless communications (visible light commu- nications – VLC) and/or acoustic communications may enable the emerging paradigm of Internet of Lighting.
Things and Internet of underwater things. Moreover, MQTT can be an easy system to describe as well as its possible applications.
- IoT-Edge-Cloud continuum essentials (15min): The balance of processing and workload in IoT network will be considered. The conventional approach, in fact, demands all the processing to the gateways while the emerging paradigm of edge/cloud/fog computing moves the focus from core network to periphery. In the context of IoT with low cost and low power energy, this paradigm should be reshaped in order to meet the main goals of long-life devices, low signaling and fresh processing [3].
- IoT security issues (25min) The IoT-Edge-Cloud shift is seen as a key shift to enable the 6G systems, but it is also with a new security landscape, that will be analysed in terms of vulnerability and impact towards the upper layers [8]. The specific issues in terms of security, related to the IoT-Edge architectures will be presented, by providing a description of the typical attacks at the different layers. Technologies as 5G will be analysed from a security viewpoint.
- The AI-aided security double-edged sword dilemma: (smart) security countermeasures and advanced at- tacks (40 min) The evolution of the security landscape based on the massive relying on Machine Learning approaches will be presented; In particular, AI techniques, like in example ML approaches, will be discussed as a mean to improve attacks prevention and detection but also to create more sophisticated and undetectable attacks. A taxonomy of the different advanced Smart Attacks will be detailed [4].
- Use cases (10 min) An IoT-Edge architecture based on heterogeneous devices will be introduced. The use of ML for malicious attacks and to counteract a target illicit device will be presented with experimental results, to demonstrate the double nature of ML in the security domain.
