The subject of this thesis is part of the context of (cyber)criminal activity detection and countermeasures by developing a surveillance infrastructure for target areas. This infrastructure consists of heterogeneous nodes, where the deployment of reprogrammable connected objects is envisaged, as well as thin and flexible planar antennas made of conductive nanomaterials. Furthermore, the nodes can intercept and counter specific target signals, minimizing the impact on other neighboring nodes, which communicate on the same frequencies. The reason why connected objects are considered heterogeneous is related to the fact that in a hyperconnected world, we may be in the presence of numerous devices communicating and emitting electromagnetic waves at overlapping frequencies. On the one hand, we hope to be able to consider specifically designed antennas that can blend into the environment to exploit data captured in the environment, which may come from people (i.e., "passive" interactions with the antennas) or from connected objects, which emit signals that are captured by the antennas. On the other hand, we also hope to use reprogrammable connected objects that can coexist and interact with the "thin" antennas.

Different scenarios will be considered, distinguishing the passive case where there are emissions in a certain area, but there are no connected objects communicating.

Another scenario is based on the presence of connected objects that emit signals that will be intercepted and analyzed. The environments can be indoor or outdoor, characterized by significantly different interactions and phenomena regarding the signals [bel24][alla-wisec24]. In a scenario without connected objects, the goal will be to leverage the surveillance infrastructure to detect the presence of signals.

In a scenario with connected objects, we envision collecting raw data/signals/I/Q, processing them, and analyzing them. In both cases, we will develop learning approaches to characterize the environment. For transmitting devices, the goal is to identify: -) the communication technology (i.e., 3G/4G/5G, Bluetooth, WiFi), -) identify the frequencies in which the target devices are transmitting, -) perform real-time demodulation, -) identify the devices, -) locate the devices, and -) intercept the signals from the target devices.

[alla-acsac24] Ildi Alla, Selma Yahia, Valeria Loscri, Hossien Eldeeb. Robust Device Authentication in Multi-Node Networks: ML-Assisted Hybrid PLA Exploiting Hardware Impairments. Annual Computer SecurityApplications Conference (ACSAC), Dec 2024, Waikiki, Hawaii, USA, United States. hal-0472749

[alla-wisec24] Ildi Alla, Hervé B. Olou, Valeria Loscri, and Marco Levorato. 2024. From Sound to Sight: Audio-Visual Fusion and Deep Learning for Drone Detection. In Proceedings of the 17th ACM Conference on Security and Privacy in Wireless and Mobile Networks (WiSec '24). Association for Computing Machinery, New York, NY, USA, 123–133. https://doi.org/10.1145/3643833.3656133

[bel24] M. Belhor, A. Savard, A. Fleury, P. Sondi and V. Loscri, "Enhanced RF-based 3D UAV Outdoor Geolocation: from Trilateration to Machine Learning Approaches," 2024 IEEE 27th International Symposium on Real-Time Distributed Computing (ISORC), Tunis, Tunisia, 2024, pp. 1-8, doi: 10.1109/ISORC61049.2024.10551331.

Funding category: Contrat doctoral



PHD title: Doctorat in Mathematics and Digital Science

PHD Country: France