Peer-to-peer network modeling for adversarial proactive cyber defenses

Abstract

This thesis implements a novel peer-to-peer network simulator that integrates co-evolutionary algorithms in order to model adversarial attack and defense dynamics in networks. Modeling this behavior is desirable as it allows for network designers to better develop network defense strategies against adaptive cyber attackers. By developing a network simulator that implements a peer-to-peer protocol, we were able to control the environment and abstract away many of the complex details that would normally arise from using a live network. Because of this environment, we were able to design attack and defense models and grammars, construct arbitrary network topologies, and rapidly test adversarial behavior using the integrated coevolutionary algorithms. Second, the thesis implements the integration of the coevolutionary algorithms with a more complex, proprietary emulator that implements an advanced version of Chord. Our experiments with this system start to investigate the effectiveness of peer-to-peer networks as defenders as well as elucidate the issues of integrating coevolutionary algorithms in a real-world system.