Extracting Length Field of Unknown Binary Network Protocol from Static Trace
Xiuwen Sun, Zhihao Wu, Jing Lin, Pengfei Fu, Jie Cui, Hong Zhong · 2023
Network protocol specification is essential in analyzing and evaluating network functionality, performance, and security. However, increasing private protocols become a hindrance to these features. The existing works study how to extract protocol keyword fields rather than infer the semantics of the fields, such as the length field, which can indicate the length associated with a message and is fundamental for deep analysis of network protocols. In this paper, we propose a nonparametric and unsupervised method, ROSE, to extract the length field of unknown binary network protocols from static traces. It segments the fields from the raw network trace and gets the inferred length of a subset of messages by clustering similar fields with k-means. Then, it generates candidate fields using n-gram and builds a multidimensional equation based on the length of the clustered messages and the candidate length fields. Finally, ROSE extracts the inferred length fields through linear regression. As far as we know, it is the first study on extracting length field from the static trace. The evaluation experiments using raw network traces exhibit high precision and recall in extracting the length field or identifying protocols without the length field.