Thesis in Information Theory and Machine Learning “Security in Distributed Detection”
Thèse posted 3 mois il y aDescription
The Thesis Topic:
In the future, billions of IoT devices and sensors will be connected and cooperate together to detect events in distributed monitoring and alert systems in applications like intrusion detection, target tracking, smart grids and smart homes, industry 4.0, e-Health or intelligent transportation (connected cars or drones). Many of these applications have system-critical security requirements, in the sense that eavesdroppers and intruders should not learn or influence decisions or data and measurements.
The objective of this thesis is to characterize the fundamental limits and devise new strategies of distributed detection systems with security constraints against external and internal eavesdroppers that should not be able to learn the data and measurements performed at the sensors. In this goal we shall propose an information-theoretic analysis as well as learning-assisted detection strategies. To derive the fundamental limits of the described systems, we shall rely on a recent converse strategy that we pioneered in by using a change-of-measure argument and by proving asymptotic Markov chains. Leveraging on these limits and with the aid of deep learning and reinforcement learning techniques, smart device selection strategies and practical secure distributed detection systems will be proposed.
Supervisors: Mireille Sarkiss, Telecom SudParis, IP Paris, mireille.sarkiss@telecom-sudparis.eu and Michèle Wigger, Telecom Paris, IP Paris, michele.wigger@telecom-paris.fr
Working Environment: The PhD student will be enrolled at the doctoral school of IP Paris, and be working in Palaiseau, (19, place Marguerite Perey) in the building shared by Telecom Paris and Telecom SudParis. Regular one-to-two weekly meetings will be held with both supervisors. Participation in biweekly research group meetings is excepted, as well as participation in international workshops and schools.
Funding: The PhD will be funded by the PEPR 5G project.
Research Domains: The PhD topic lies in the fields of information theory, machine learning, and cybersecurity.