Cours/Séminaire
Notice
Lieu de réalisation
Paris
Langue :
Anglais
Crédits
Mireille Sarkiss (Intervention)
Conditions d'utilisation
Droit commun de la propriété intellectuelle
DOI :
10.60527/sbrk-5c80
Citer cette ressource :
Mireille Sarkiss. FOUNDS. (2024, 28 juin). FOUNDS SEMINAR SERIES TALK 10 - Distributed Binary Hypothesis Testing. [Vidéo]. Canal-U. https://doi.org/10.60527/sbrk-5c80. (Consultée le 25 avril 2025)
FOUNDS SEMINAR SERIES TALK 10 - Distributed Binary Hypothesis Testing
Réalisation : 28 juin 2024
- Mise en ligne : 2 juillet 2024
- document 1 document 2 document 3
- niveau 1 niveau 2 niveau 3
Descriptif
- Title: Distributed Binary Hypothesis Testing
- Abstract: Distributed binary hypothesis testing (DHT) problem has many applications in IoT and future networks such as distributed decision and surveillance systems for security, health monitoring, intelligent car control, etc. In these systems, devices and sensors collect some data to help the decision centers distinguish between the normal situation (null hypothesis) and the alert situation (alternative hypothesis). The hypotheses are assumed to determine the joint probability distribution underlying the data observed at the various nodes. The objective of such problems is to maximize the exponential decay of miss-detection (type-II error) probabilities while preserving the false alarm (type-I error) probabilities below given thresholds, referred to as the Stein exponent by Ahlswede and Csiszar. Yet, security is a critical concern in these systems, in the sense that eavesdroppers and intruders should not be able to learn the data or decisions or even to detect the presence of communications. In this talk, we focus on a DHT problem with a single sensor, single decision center, and an eavesdropper, all having their own source observations. We consider first testing against independence over a rate-limited noiseless. We characterize the largest possible type-II error exponent at the legitimate receiver under constraints on the legitimate receiver's type-I error probability and the equivocation (uncertainty) measured at the eavesdropper about the sensor's observations. Then, we consider the DHT over a discrete memoryless channel under the stronger covertness constraint. Thus, we impose in this case that an eavesdropper should not be able to determine whether communication is ongoing or not. The main result here is an upper-bound on the largest possible Stein exponent showing that it cannot exceed the largest exponent achievable under zero-rate communication over a noise-free link.