Recovering Models of Network Protocol Using Grammatical Inference

Abstract

In this paper, a novel solution to induce a concise automata representation for network protocol is presented and evaluated. First, the feature element sequences representing the protocol interacting process are extracted. Second, a partial automata that accepts these sequences is aggregated; Then, the automata are labeled using a specific heuristic based on the network characteristic; Last, a more generalized and minimal automata is generated by the equivalent state merging algorithm. The models induced are applied to a number of real-world applications that implement various protocols. The evaluation results demonstrate that our techniques presented are valid.