Quantifying IIoT Sensor Node Criticality by Fusing its Data Criticality and Security Vulnerability

1 Introduction

• IIoT sensors support real-time monitoring, process optimisation, and predictive maintenance, while their reliability, security, and effectiveness remain challenges.The introduction links these functions to product quality control and operational efficiency.Source: §1 · p. 1
• Existing attack- or vulnerability-based methods do not account for the criticality of sensor data.Data-criticality research, in turn, does not account for cybersecurity threats.Source: §1 · p. 2
• The paper proposes quantifying IIoT sensor node criticality by fusing data criticality and security vulnerabilities with Dempster-Shafer theory.The introduction identifies the absence of methods combining these factors as the research gap.Source: §1 · p. 2
• The proposed framework evaluates fused criticality using Belief and Plausibility scores and compares the rankings with pre-fusion data-criticality and security-vulnerability rankings.The evaluation uses red-wine production data.Source: §1 · p. 2

2.1 Assessment of Critical Assets in the IIoT-Enabled Digital Systems

• Critical assets include information, data, devices, and equipment whose disruption can affect product quality control and productivity.Assets may be classified by their importance to production and quality assurance, with security controls selected accordingly.Source: §2.1 · pp. 2–3
• Assessing critical data and associated risks supports protection of sensitive equipment and business-critical systems in digital manufacturing.The section links risk classification to understanding exposed vulnerabilities and their effects on assets.Source: §2.1 · p. 3

2.2 Vulnerability Assessment in the IIoT-Enabled Digital Systems

• Vulnerability assessment research addresses weaknesses in IIoT systems, including hardware, software, applications, and devices, using approaches such as attack-path analysis.The cited SCADA assessment draws on field incident responses; older devices may be exposed to malware and high network traffic.Source: §2.2 · pp. 3–4

2.3 Assessment of IIoT Sensor Node Criticality

• Connected IIoT sensors and devices expand the number of potential attack endpoints, making their vulnerabilities and operational risks important to assess.Risks include effects on process safety, reliability, environmental factors, and data confidentiality, integrity, and availability.Source: §2.3 · p. 4

3 Literature Review

• The review groups sensor-node criticality methods into cyber exploits, data criticality, and node failures or other factors.It presents sensor nodes as important to data collection, communication, production, and product quality.Source: §3 · p. 4

3.1 IIoT Node Criticality Based on Cyber Exploits

• Cyber-exploit approaches include vulnerability scoring with CVSS, but methods developed for consensus networks are not directly applicable to independently operating IIoT sensors.The review also notes that network-level cyber-risk assessment is not directly applicable to sensor vulnerability assessment.Source: §3.1 · pp. 4–5

3.2 IIoT Sensor Node Criticality Based on Data Criticality

• Data-criticality assessment evaluates how sensor data and component variation relate to product quality and ranks sensor nodes by that significance.The cited prior method examines quality sensitivity and the percentage of quality variation associated with component changes.Source: §3.2 · p. 5

3.3 Criticality Based on Node Failures and Others

• Failure- and topology-based approaches assess criticality through cascading failures, connectivity, load, or network structure, but may not capture sensor-specific impacts on quality.A load-based K-shell method is described as focused on networking devices such as gateways and edge nodes.Source: §3.3 · pp. 5–6

3.4 Motivation of the Research

• The authors identify the need for a unified assessment integrating data significance with cybersecurity risks to evaluate IIoT sensor-node criticality.They frame this motivation in terms of protecting manufacturing operations and product quality.Source: §3.4 · p. 6

4 Proposed Approach of Quantifying IIoT Sensor Node Criticality

• The framework assesses IIoT sensor-node importance by considering data criticality and sensor security vulnerabilities.The section describes this as an approach for a digital manufacturing setup.Source: §4 · p. 6
• The approach is demonstrated in a wine-manufacturing context.The passage describes the setting as leveraging IIoT integration and advanced data analytics for high-quality wine production.Source: §4 · p. 6
• The framework is presented in relation to high-quality wine production and advanced data analytics.Source: §4 · p. 6

4.1 Overview of the Proposed Approach

• Blocks 1–5 collect sensor data and vulnerability information, assess data criticality and security-vulnerability likelihoods, then fuse those probabilities into node criticality.Data collection and vulnerability information are handled separately before their assessed probabilities are fused.Source: §4.1 · pp. 6–7
• Block 6 ranks node criticality from 1, the highest, to k, the lowest, for application systems’ decision-making.Application systems monitor or access these rankings.Source: §4.1 · p. 7
• The approach assumes more-critical nodes have more important data and are more likely to be compromised; Cloud or Edge Systems execute the process.Source: §4.1 · p. 7

4.2 Assessing the Probability of Sensor Data Criticality

• A sensor’s criticality probability rises when its data is critical because manufacturing product quality may be significantly affected.Source: §4.2 · p. 7
• Data importance is assessed by its value for decision-making or system functionality.Source: §4.2 · p. 7
• The data-criticality rank Ri1 and sensor count n determine mi1(ζ), the probability that sensor i is critical considering its data criticality.ζ denotes the observed data or vulnerability level being critical.Source: §4.2 · p. 7

4.3 Assessing the Probability of Sensor Security vulnerability

• CVSS scores range from 0 to 10, with severity bands of Low (0.1–3.9), Medium (4.0–6.9), High (7.0–8.9), and Critical (9.0–10.0).Source: §4.3 · p. 7
• CVSS vulnerability assessment is used to gauge the probability that a sensor node is critical for security and functionality.Source: §4.3 · p. 7
• Ri2 denotes sensor i’s security-vulnerability rank, and mi2(ζ) its probability of being critical for cybersecurity.Source: §4.3 · p. 7

4.4 Fusing Data Criticality and Cybersecurity Vulnerability

• Dempster–Shafer theory is selected to fuse data criticality and cybersecurity vulnerability; it supports propositions using temporal data and is commutative and associative.The passage notes its use across industrial applications and decision-making.Source: §4.4 · pp. 7–8

4.4.1 Dempster-Shafer Evidence Theory

• Basic probability assignment maps sets in the power set to masses; belief sums subset masses, while plausibility sums masses for sets intersecting the proposition.Belief and plausibility form lower and upper bounds; under normalized assignments they are related through the proposition’s complement.Source: §4.4.1 · p. 8

4.4.2 Dempster’s Rule of Combination

• Dempster’s rule combines independent belief-function evidence using a conjunctive AND operation, with conflict mass k and normalization factor 1 − k.The combination is defined for non-null bpa intersections.Source: §4.4.2 · pp. 8–9

4.4.3 Frame of Discernment

• The discernment frame distinguishes observed data or vulnerability levels as critical or not critical, with separate evidence sources for the two assessment dimensions.Source: §4.4.3 · p. 9
• The method uses an entropy-based uncertainty mass and a min() constraint to prevent negative non-critical mass before deriving lower and upper criticality limits.The lower limit is Bel_i(ζ), and the upper limit is Pl_i(ζ).Source: §4.4.3 · p. 9

4.5 Limitations and Biases

• The assessment covers only a limited selection of sensors monitoring wine production.The authors note that criticality and security vulnerability were computed for this selected set.Source: §4.5 · p. 9
• Subjective definitions and measurements of data criticality and security vulnerability leave room for human error and bias.Weight assignment also involves subjective judgment, and stakeholders with different priorities may produce varying criticality assessments.Source: §4.5 · pp. 9–10
• Assumptions and limitations of D-S Theory could affect the accuracy and reliability of the findings.Source: §4.5 · p. 9
• Generalising the results to all IIoT sensor nodes may not be feasible because criticality and vulnerability vary by application and environment.Source: §4.5 · p. 10
• Using publicly available information to assess security vulnerability may not accurately capture potential threats, compromising the overall criticality assessment.The authors question the validity and applicability of the vulnerability metrics used in their computations.Source: §4.5 · p. 10

5.1 Case Scenario

• The wine scenario uses IIoT sensors to monitor production parameters, and manipulated sensor communications could affect fermentation, blending, or bottling decisions and wine quality.The monitored parameters include acidity, sugar, chloride, sulphate, SO2, pH, density, and alcohol.Source: §5.1 · p. 10

5.2 Dataset - Evidences for Fusion

• The red-wine dataset contains 11 independent characteristics paired with wine-quality ratings, with sensors S01–S11 measuring the respective characteristics.The characteristics include fixed acidity, volatile acidity, citric acid, residual sugar, chlorides, free and total sulphur dioxide, density, pH, sulphates, and alcohol.Source: §5.2 · p. 11

5.2.1 IIoT Sensor-Generated Critical Dataset

• Data criticality is assessed using characteristic–quality correlation, percentage change in quality, and quality sensitivity, then combined as weighted, normalised rankings.The resulting parameter criticality values are added to the fusion dataset.Source: §5.2.1 · p. 11

5.2.2 IIoT Sensor Security Vulnerability Dataset

• The vulnerability dataset uses CVSS metrics tailored to wine sensors, with impact linked to characteristic–quality correlations and threats assessed for the system’s network and cloud setting.The described architecture includes a shared WiFi network, a central PLC, and public cloud storage.Source: §5.2.2 · p. 11

5.2.3 Probabilities and Uncertainties Associated with the Evidence

• Equations 1–2 calculate evidence probabilities and Equations 11–12 calculate their uncertainties; the resulting probabilities and uncertainties are reported in Table 1.The evidence covers sensor-generated data criticality and sensor security vulnerabilities.Source: §5.2.3 · p. 11

5.3 Sensor Node Criticality Fusing Data Criticality and Security Vulnerability

• The fusion framework combines data criticality and vulnerability evidence, and Table 2 compares fused prioritisation with data-criticality and CVSS 4.0 and 3.1 vulnerability rankings.Set-A uses CVSS 4.0 vulnerabilities, while Set-B uses CVSS 3.1 vulnerabilities.Source: §5.3 · pp. 11–12, Table 2 · p. 12

5.3.1 Fusing Data Criticality and CVSS 4.0 Security Vulnerability

• Set-A combines sensor-generated data criticality with vulnerabilities scored using CVSS 4.0, then applies Dempster’s rule to account for evidence probabilities and uncertainties.The resulting belief scores range from 0.5465 to 0.9835, and plausibility scores range from 0.8588 to 1.0.Source: §5.3.1 · p. 12, Figure 2 · p. 12

5.3.2 Fusing Data Criticality and CVSS 3.1 Security Vulnerability

• Set-A belief scores span 0.5465–0.9835 and Set-B scores span 0.1992–0.9834; both identify S02, S11, S04, S01, and S07 as especially critical.The figures represent belief as the lower probability limit and plausibility as the upper limit.Source: §5.3.1 · p. 12, Figure 2 · p. 12, §5.3.2 · p. 13

5.4 Comparative Analysis of the Results

• Top-five rankings from sensor-generated data align with fused rankings for both sets, while vulnerability-based rankings differ between CVSS 4.0 and CVSS 3.1.Source: §5.4 · p. 14
• Bel scores range from 0.5465 to 0.9835 for Set-A versus 0.1992 to 0.9834 for Set-B; Set-A is better supported, while Set-B varies more.Set-A also has slightly higher overall Pl scores.Source: §5.4 · pp. 13–14, §5.5 · p. 14
• Pl scores for both evidence sets range from 0.86 to 1.0, with Set-A slightly higher overall.Source: §5.5 · p. 14, §5.4 · p. 14

5.5 Summary of Results

• S07, S01, S04, S11, and S02 are highly critical, corresponding respectively to TSD-, FA-, RS-, ALC-, and VA-Sensors in digital wine manufacturing.The authors state these sensors are crucial for wine quality and require enhanced security measures.Source: §5.5 · p. 14

• The framework combines security vulnerability analysis and data criticality metrics to rank IIoT sensor-node criticality.Source: §6 · p. 14
• The ranking is intended to identify high-priority nodes for cybersecurity enhancement and help industries allocate security resources.Source: §6 · p. 14
• Future work could expand the approach’s applicability and explore dynamic vulnerability assessment for evolving threats.Source: §6 · p. 14