TRSCNN: two-dimensional protocol reverse parsing method

作者：HJ Xu, Yuqing Lan · 年份：2026 · DOI：10.1117/12.3118315 · 研究领域：Computer science、Data mining、Distributed computing、Artificial intelligence

Protocol Reverse Engineering (PRE) is a critical research topic in the fields of the Industrial Internet of Things and Industrial Control Systems (ICS). The widespread use of private and undisclosed communication protocols in industrial devices results in numerous non-standard protocols within ICS, posing significant challenges to device access, system integration, and security analysis. Meanwhile, IoT devices often employ customized protocols without public APIs, making protocol reverse engineering essential for achieving local interoperability. The primary objective of PRE is to infer field types and extract functional semantics of unknown protocols by analyzing network traffic characteristics; however, traditional approaches are limited by incomplete field descriptions and insufficient semantic inference accuracy. To address these issues, considering the temporal characteristics of industrial protocol data and the correlations among fields of the same type, we propose TRSCNN to jointly learn temporal features and spatial distribution patterns of protocol data, thereby improving the accuracy of field semantic inference. Experimental results demonstrate that our method outperforms stateof- the-art approaches in both Top-1 and Top-2 accuracy, and ablation studies further verify the complementary contributions of the temporal and spatial modules to overall performance.