We investigate the problem of modeling node capture attacks in heterogeneous wireless ad hoc and mesh networks. Gligor's work noted that classical adversarial models such as the Dolev-Yao model are known to be unsuitable for describing node capture attacks. By defining the amortized initialization overhead cost as well as the cost of capturing a node, we show that finding the node capture attack yielding the minimum cost can be formulated as an integer-programming minimization problem. Hence, there is no polynomial solution to find the minimum cost node capture attack. We show that depending on the adversary's knowledge of the constraint matrix in the integer-programming problem, different greedy heuristics can be developed for node capture attacks. We also show under what conditions privacy-preserving key establishment protocols can help to prevent minimum cost node capture attacks. Individual node storage randomization is investigated as a technique to mitigate the effect of attacks which are not prevented by the use of privacy-preserving protocols. It is shown that probabilistic heuristic attacks can be performed effectively even under storage randomization.