Physical semantic inference method for industrial control proprietary protocol data fields

doi:10.1016/j.cose.2025.104654

Physical semantic inference method for industrial control proprietary protocol data fields

2025-09-09

0

PRE

AI

DOI：10.1016/j.cose.2025.104654

Original

Original request for help

Save

Abstract

En 中文

The precise interpretation of physical semantics within industrial proprietary protocol data fields, which directly govern cyber–physical interactions, plays a pivotal role in securing industrial control systems (ICS). Current protocol reverse engineering methods face three fundamental limitations that hinder ICS security efforts: (1) discerning physical significance in raw hexadecimal streams, (2) unassisted field delineation amid unknown data types, and (3) mapping concurrent physical semantics to protocol fields without contextual references. To overcome these challenges, we propose Physeinfer, a novel physical semantic inference framework that innovatively integrates visual human-machine interface (HMI) monitoring with temporal sequence analysis. Our methodology leverages camera-acquired HMI panel differentials to extract physical semantic sequences, develops adaptive step-size strategies for longitudinal multiple sequence alignment, and employs dynamic time warping (DTW) to establish cyber–physical correlations. Evaluated in six industrial scenarios, Physeinfer was able to accurately recover the physical semantics of industrial control protocol data fields and demonstrated superior performance in field segmentation (11%–27% improvement over Netzob, MSERA, and Fieldhunter) without requiring prior knowledge of the protocol. This breakthrough establishes an essential foundation for context-aware security mechanisms in industrial infrastructure, enabling physics-informed vulnerability discovery and anomaly detection for ICS.

Physical semantic inference method for industrial control proprietary protocol data fields

Abstract

Journal

Organization