TEEM: Trust-based Energy-Efficient Distributed Monitoring for Mobile Ad-hoc Networks Andrea Lupia * , Chaker Abdelaziz Kerrache † , Floriano De Rango * , Carlos T. Calafate ‡ , Juan-Carlos Cano ‡ ,Pietro Manzoni ‡ * DIMES Dept., University of Calabria Rende (CS), Italy {alupia, derango}@dimes.unical.it † Laboratoire d’Informatique et de Math´ ematiques, University of Laghouat BP 37G, route de Ghardaia, Laghouat, Algeria a.kerrache@mail.lagh-univ.dz ‡ Department of Computer Engineering, Universitat Polit` ecnica de Val` encia Camino de Vera, S/N, Valencia, Spain {calafate, jucano, pmanzoni}@disca.upv.es Abstract—Detecting malicious and selfish nodes in mobile and wireless networks is a task that typically relies on watchdog techniques. However, such medium overhearing techniques are prone to cause a high energy consumption. In this paper we propose a distributed time division-based monitoring strategy to achieve the high security levels required while consuming less energy. Our proposal involves both trust and link duration among honest peers to fairly divide the monitoring period, and takes advantage of the periodically exchanged hello messages to make this solution fully distributed. Simulations results evidence our proposal energy efficiency, especially for high density scenarios (>100 nodes) where the consumption became stable and does not increase with the number of nodes (<4300 W·s), while ensuring high detection ratios of malicious nodes (>90%). Index Terms—Distributed monitoring; Trust; Energy consumption; MANET. I. I NTRODUCTION Solutions ensuring secure and reliable communication are typically classified into two main categories: cryptography-based, which mostly introduce a high computational delay, and trust-based solutions, that allow evaluating past interactions with less computational effort. The latter are more adapted to mobile and energy-restricted networks because they have lower computational requirements compared to the former, and so they are adopted as trust management solutions in different kinds of networks including Mobile Ad-hoc Networks (MANETs) [1] and Vehicular Ad-hoc Networks (VANETs) [2], being able to ensure high security levels with reduced computational requirements. However, despite they reduce the energy related to computation, the consumption associated to the communications interface remains the same. To overcome the aforementioned problems, and to find a proper trade-off between security levels and energy consumption, in this paper we propose a distributed trust-based monitoring solution able to face both malicious and selfish nodes while supporting node sleeping. Our proposal first combines trust evaluation and link duration to select the most trusted and stable neighbors in the neighborhood. Afterward, nodes trusting each other will mutually synchronize in a distributed manner, and agree on the monitoring schedule for their neighborhood; this is achieved through some additional fields appended to periodic ’Hello’ messages. With this technique the energy consumption is significantly reduced, meaning that the more trusted neighbors a node has, the less time it needs to dedicate to network monitoring tasks, thereby extending its lifetime. It is worth mentioning that our solution is independent and can be implemented over any routing protocol; in particular, for this work, we chose to implement it over the well-known Ad-hoc On-demand Distance Vector (AODV) routing protocol [3]. Obtained simulation results using the NS-3 simulator show that our distributed monitoring technique significantly reduces the energy consumption by more than 50%, without affecting the detection performances (>90%). The rest of the paper is organized as follows: Section II details how trust and link duration can be computed and combined together in order to select the most trusted and stable neighbors with whom the monitoring period will be shared. We explain the distributed monitoring process in Section III, and we then evaluate its performance in Section IV. Finally, Section V concludes the paper. II. TRUST AND LINK DURATION COMPUTATION A node j is considered companion of another node i if node j has a high trust value from the perspective of i, and the communications between them are mostly stable. Therefore, a node j is included in the companions table only if its Companion Score (CS) is higher than a predefined threshold; this node will be also removed from the table once its CS(i, j ) value goes below the same predefined threshold. To compute CS we use the following equation: CS(i, j )=(α · T i,j ) + [(1 - α) · LSI(i, j )] (1) 978-1-5090-5856-3/17/$31.00 ©2017 IEEE 133