Abstract
Data classification and grading management is crucial for ensuring data security and facilitating rational development and utilization. Petrochemical enterprises, as custodians and processors of data, urgently need to establish a unified and scientific system for safeguarding classified data. This paper proposes a comprehensive framework for classifying and grading data within petrochemical enterprises to meet the stringent requirements of China’s recent data security laws. By combining hierarchical "line" classification with flexible "plane" classification, the authors establish a four-tiered grading system based on the potential impact of data breaches on national security, corporate interests, and individual privacy.
Introduction
As the industrial Internet matures, petrochemical enterprises find themselves managing massive, highly specialized, and sensitive datasets. While laws like the Data Security Law of the People’s Republic of China (2021) mandate data protection, the petrochemical sector lacks a unified, scientific standard for how to actually categorize this information.
Currently, many enterprises face inconsistent standards between internal departments and external partners. Without a clear "map" of what data is sensitive versus what is public, it is nearly impossible to implement efficient security policies or facilitate safe data sharing. This research aims to bridge this gap by providing a standardized methodology tailored to the unique business logic of the oil and gas industry.
Methodology: A Hybrid Classification Approach
The proposed framework utilizes a hybrid classification approach that balances structure with flexibility, drawing upon existing Chinese national standards and industrial guidelines (such as GB/T 10113-2003 and Q/SY 10345-2020).
- Line Classification: A traditional hierarchical approach where data is nested in levels (e.g., Finance -> Payroll -> Hourly Rates). This is effective for showing logical relationships.
- Plane Classification: A multi-dimensional approach where data is categorized by independent attributes (e.g., origin, business domain, security level), allowing for more parallel and flexible management.
Data Resource Inventory and Classification
The framework divides the data universe into two primary domains: the Public Data Subject Domain and the Business Subject Domain. The Public Data domain includes reusable data requiring high consistency across the company, such as equipment master data. The Business domain covers data specific to professional fields like exploration, production, or refining, typically divided into no more than six levels to maintain clarity.
As shown in the figure above, the inventory structure (L1 through L2 and beyond) allows for a granular breakdown of assets. For instance, "Production Control" (L1-2) branches into specific operational areas like "Exploration and Production" (L2), which further subdivides into specialized engineering and operational data points.
Data Grading and Impact Assessment
For grading, the paper introduces a matrix based on three critical factors: the Objects of Impact (National security, the petrochemical sector, the enterprise, or individuals), the Scope of Impact (Multiple sectors, a single enterprise, or specific internal departments), and the Extent of Impact (Severe, serious, or general).
This results in a four-level grading system:
- Grade 4 (Top): Breaches impacting national security or multiple sectors with severe/serious damage. Following national law, these are treated as "Core Data."
- Grade 3 (High): Breaches impacting the enterprise's overall interests or secondary bodies with serious damage.
- Grade 2 (Medium): Breaches impacting internal departments or individual privacy with serious/general damage.
- Grade 1 (Low): Breaches with general impact on individuals or tertiary bodies, resulting in low importance.
Implementation and Future Work
The authors provide a three-stage workflow for implementation: domain subdivision (identifying sources), data classification (assigning business categories), and grading determination (assessing impact).
One key insight is that security levels are not static. Factors like "significant aggregation"—where combining many non-sensitive data points creates a sensitive set—or changes in timeliness can cause a data grade to fluctuate. Furthermore, manual classification is insufficient for modern industrial scales; the authors advocate for automated software tools using techniques like automated labeling to handle batch classification.
While the framework provides a robust starting point, the authors acknowledge that there is currently no industry-wide consensus on what specifically constitutes "important data" inventories. Future work will involve refining these methods through pilot projects across various sub-sectors of the petroleum and natural gas industry to move toward a more unified scientific method and a robust data security ecosystem.
Summary updated