Evolutionary Design of Attack Strategies

Abstract

We propose and investigate a new concept for automatic search for attack strategies and demonstrate its use in the area of wireless sensor networks. We exploit mechanisms inspired by the biological evolution, known as Evolutionary Algorithms and we use a combination of simulation (or real system execution) with a candidate attacks generator. Each candidate is evaluated, its success is measured and used as a guiding input for the next generation of candidates. Evolutionary Algorithms perform well in case they evaluate a large number of candidate attack strategies, and thus we have focused on applications where quality can be assessed fast (in order of seconds and optimally less).