Attackers in wireless sensor networks will be neither random nor jumping – Secrecy amplification case

Abstract

Partially compromised network is a pragmatic assumption in many real-life scenarios. Secrecy amplification protocols provide a significant increase in the number of secure communication links by reestablishing new keys via different communication paths. Our paper shows that research in the area of secrecy amplification protocols for ad-hoc networks has been based on rather simplified foundations w. r. t. attacker models. The attacker does not behave randomly and different attacker capabilities and behaviour have to be considered. We provide means to experimental work with parametrisable attacker capabilities and behaviour in realistic simulations, and evaluate the impact of the realistic attacker properties on the performance of major amplification protocols (Full details, paper supplementary material and source codes can be found at http://crcs.cz/papers/cans2016.). We also show which secrecy amplification protocols perform best in different attacker settings and help to select a protocol that exhibits good results in a prevalent number of inspected scenarios.