Automatic Identification of Industrial Control Network Protocol Field Boundary Using Memory Propagation Tree

The knowledge of protocol specification, especially protocol field boundary, is invaluable for addressing many security problems, such as intrusion detection. But many industrial control network (ICN) protocols are closed. Closed protocol reverse engineering has often been a time-consuming, tedious and error-prone process. Some solutions have recently been proposed to allow for automatic protocol reverse engineering. But their prerequisites, e.g. assuming the existence of keywords or delimiters in protocol messages, limit the scope of their efforts to parse ICN protocol messages. In this paper, we present AutoBoundary that aims at automatically identifying field boundaries in an ICN protocol message. By instrumenting and monitoring program execution, AutoBoundary can obtain the execution context information, and build a memory propagation (MP) tree for each message byte. Based on the similarity between MP trees, AutoBoundary can identify protocol field boundaries, automatically. The intuition behind AutoBoundary makes it suitable for ICN protocols, which have the characteristics of no delimiter, no keyword, and no complex hierarchical structure in the message. We have implemented a prototype of AutoBoundary and evaluated it with 62 ICN protocol messages from 4 real-word ICN protocols. Our experimental results show that, for the ICN protocols whose fields are byte-aligned, AutoBoundary can identify field boundaries with high accuracy (100% for Modbus/TCP, 100% for Siemens S7, and 94.7% for ISO 9506).