Section: New Results

Wireless Networks

Participants : Eitan Altman, Philippe Nain, Giovanni Neglia.

Estimation of population sizes in sensor networks

We have been working on several problems related to the estimation of population sizes. In collaboration with D. Kumar (IBM Research Center, Hawthorne, USA) and T. Başar (Univ. of Illinois at Urbana-Champaign, USA), E. Altman develops in [73] a Wiener filter that allows to estimate the number of sensors that cover the space at some selected points. The authors take advantage of spatial correlations between the number of sensors covering different points in order to derive the filter. We note that causality is not an issue in space, in contrast to filtering at different points in time.

In collaboration with A. Ali, T. Chahed and M. K. Panda (Telecom SudParis, France), D. Fiems (Gent Univ., Belgium), and L. Sassatelli (I3S, Univ. Nice Sophia Antipolis - CNRS, France), E. Altman has used in [37] Kalman filtering theory in order to estimate the number of mobiles in a delay tolerant ad-hoc network which have a copy of a broadcasted message.

Cellular networks: Small cells

Analysing performance measures of cellular systems combines tools from queueing theory and stochastic processes, on one hand, and geometric considerations on the other hand. In [72] , V. Kavitha (Mymo Wireless, Bangalore, India), S. Ramanath (Lekha Wireless Solutions, Bangalore, India), and E. Altman compute the time it takes to transmit a file taking into account the channel conditions which vary due to mobility of terminals. Mobility considerations play a key role in small cells since handover may occur way before the transmission of the file ends.

Multi scale fairness concepts for resource allocation in wireless networks

In many applications that require resources, one needs these resources within some given deadline. These impose constraints when attempting to allocate resources fairly. In [14] , E. Altman, K. Avratchenkov and S. Ramanath have extended the $\alpha$ fairness concept by Mo and Walrand so as to include time constraints. They study the question of how to compute such constrained fair allocation, and derive some asymptotic properties of constrained fair assignment.

Self organization in cellular communications

Self organization is an approach to design networks so as to allow them to configure in an automatic way. This allows to reduce the complexity in systems containing thousands of mobiles and a huge number of small cells. In cellular networks, self organization can be used for deciding on time or frequency reuse according to the interference in these time and frequency slots from other cells. The impact of self organization on communications are derived in [55] and [21] by R. Combes, and Z. Altman (Orange Labs, Issy les Moulineaux), in collaboration with E. Altman.

Streaming over wireless

In [75] , E. Altman and M. Haddad study in collaboration with T. Jiménez and R. El-Azouzi (Univ. Avignon/LIA) and S.-E. Elayoubi (Orange Labs, Issy les Moulineaux) streaming service over cellular networks. The purpose is to obtain the exact distribution of the number of buffer starvations within a sequence of $N$ consecutive packet arrivals. This is then applied to optimize the quality of experience (QoE) of media streaming service over cellular networks by exploiting the tradeoff between the start-up delay and the starvation.

Wireless network security

The operation of a wireless network relies extensively on exchanging messages over a universally known channel, referred to as the control channel. The network performance can be severely degraded if a jammer launches a denial-of-service (DoS) attack on such a channel.

In [94] , P. Nain, M. Krunz, H. Rahbari and M. J. Abdel Rahman (all three from Univ. of Arizona, USA) design frequency hopping (FH) algorithms that mitigate DoS attacks on the control channel of an asynchronous ad hoc network. More specifically, three FH algorithms (called NUDoS, KMDoS, and NCMDoS) are developed for establishing unicast (NUDoS) and multicast (KMDoS and NCMDoS) communications in the presence of multiple jammers. KMDoS and NCMDoS provide different tradeoffs between speed and robustness to node compromise. These algorithms are fully distributed, do not incur any additional message exchange overhead, and can work in the absence of node synchronization. Furthermore, KMDoS and NCMDoS have the attractive feature of maintaining the multicast group consistency. NUDoS exploits the grid quorum system, whereas KMDoS and NCMDoS use the uniform k-arbiter and the Chinese remainder theorem (CRT) quorum systems, respectively. Extensive simulations are used to evaluate these algorithms.