A Hierarchical Clustering Based Feature Word Extraction Method

Abstract

Feature words extraction is an important task in protocol reverse engineering. How to simplify the feature word extraction process and how to increase the recall ratio remains a considerable issue. In this paper, we propose a hierarchical clustering based approach to extract feature words from unknown protocols. The approach does not need any prior knowledge of the unknown protocols. Feature words are extracted on the basis of hierarchical clustering and longest common sequences. We cluster the similar sequences, and then extract their longest common subsequences. Finally, we extract feature words from longest common subsequences through position information of every byte. Experimental results show that the proposed approach can reach a better recall result compared with traditional methods.