Lamoussa Sanogo

Internet of Things (IoT) devices are highly vulnerable in terms of cybersecurity. Vulnerability factors include "(1)-weak security design, which focuses more on services than on security, (2)-wireless communication makes attacks such as replay, mimicking, man-in-the-middle and spoofing easier, (3)-interaction with devices through firmware updates, as well as collecting and sharing information for economic purposes". Attacks targeting IoT will certainly increase with the emergence of Artificial Intelligence (AI). In this context, securing IoT devices remains a major challenge, because of their resource-constrained nature, the high decentralization of IoT networks as a result of direct machine-to-machine communication, making it difficult, for example, the deployment of firewall-like solutions, and the use of a multitude of protocols in IoT, making protocol independency a requirement for security solutions. The objectives of this thesis are to propose original security solu! tions at the physical layer to allow IoT devices to discover their environment, integrate into it with confidence, and adapt to it depending on the number of devices, the required data rate, the required range and transmission power, and the available energy. This adaptability to the environment must take into account the mobility of these devices, which must be able to belong to different networks at different moments, as well as the trust that can or cannot be placed in these different networks. New security mechanisms that take into account IoT-related constraints need to be found: security solutions with a lightweight memory footprint, low computational workload, energy-efficient, generic and protocol-independent. In the case of successful attacks, countermeasures through hardware/software reconfiguration and readaptation techniques are also needed. After highlighting the limitations and opportunities of fingerprinting, through a hands-on application using Power Spectra! l Density (PSD) as fingerprinting feature, this work proposes a new and innovative security approach, through dynamic authentication of messages using a technique based on Polarization Shift Keying. This security approach - lightweight, protocol-independent and non-invasive - consists in cryptographically controlling the polarization of the radio wave emitted by a device in order to transmit, in parallel with the main message, authentication data for this main message. The experimental results of this new authentication method show that this technique is a promising approach for securing communications in IoT based on the PHY layer. The proposed solutions are generic and protocol-independent.

Email:

Address:Toulouse, Midi-Pyrenees, France