Inferring Military Activity in Hybrid Networks through Cache Behavior

In this work, we consider a tactical hybrid network where a military adhoc network utilizes a commercial cellular network as backbone for data traffic. It is assumed that all traffic goes through a cache at the cellular base station with Least-Recently Used (LRU) policy. We study the possibility of inferring the military traffic rate from cache hit/miss observations of a single adversary node. We show the possibility of such an inference by presenting algorithms for estimating the parameters of the hybrid network. First, we show that an adversary node can estimate the cache and workload parameters of a hybrid network by observing the hit/miss results of her queries. We then show that these estimates can be used to infer the military traffic rate.