The DOI matches its Crossref record; 14 of the 14 references it cites were found in a registry or an index; 12 of the 39 sources its claims cite were checked for support.
Sensor-Based Cyber Risk Management in Railway Infrastructure Under the NIS2 Directive
Open accessAbstract
This study introduces a sensor-centric cybersecurity framework for railway infrastructure that extends Failure Mode and Effects Analysis (FMEA) from traditional reliability evaluation into the domain of cyber-induced failures affecting data integrity, availability and authenticity. The contribution lies in bridging regulatory obligations of the NIS2 Directive with field-layer monitoring by enabling risk indicators to evolve dynamically rather than remain static documentation artefacts. The approach is demonstrated using a scenario-based dataset collected from approximately 250 trackside, rolling-stock, environmental and power-monitoring sensors deployed over a 25 km operational segment, with representative anomalies generated through controlled spoofing, replay and injection conditions. Risk was evaluated using RPN scores derived from Severity–Occurrence–Detectability scales, while anomaly-detection performance was observed through detection-latency variation, changes in RPN distribution, and qualitative responsiveness of timestamp-based alerts. Instead of presenting a fixed benchmark, the results show how evidence from real sensor streams can recalibrate O and D factors in near-real-time and reduce undetected exposure windows, enabling measurable compliance documentation aligned with NIS2 Article 21. The findings confirm that coupling FMEA with streaming telemetry creates a verifiable risk-evaluation loop and supports a transition toward continuous, evidence-driven cybersecurity governance in railway systems.
Overview
AI snapshot
Written from the paper’s full text.
TL;DR
The paper proposes a sensor-centric FMEA framework for railway cyber risk management aligned with NIS2 Article 21, demonstrated on approximately 250 sensors over a 25 km operational segment with controlled spoofing, replay, and injection scenarios. It reports scenario- and expert-based improvements including 45% detection accuracy improvement, 60% detection latency reduction, and 34% mean RPN reduction, while mapping risk indicators to NIS2 Criteria 1 and 6.
Key findings
- The adapted FMEA extends traditional reliability analysis to cyber-induced sensor failures affecting data integrity, availability, and authenticity; RPN is calculated as S × O × D, with mean RPN 326 and top quartile exceeding 430, classified as unacceptable under the adopted scale.
- Risk distribution across NIS2 criteria was approximately 42% for Criterion 1, 36% for Criterion 6, and 22% shared/cross-domain; high-risk modes were primarily associated with compromised data integrity and insufficient monitoring.
- Detectability was the dominant factor in RPN escalation: a ±1 change in D led to an average ±12–18% change in total RPN, and scenarios with O or D ≥ 8 automatically pushed RPN into the “unacceptable” tier.
- Scenario/expert-based results report 45% improvement in detection accuracy, 60% decrease in detection latency, and 34% reduction in mean RPN; one signaling example reduced detection latency from 10 s to 2 s, and one firmware-validation example reduced RPN from 405 to 270.
Methods
- Case study on a representative railway IoT architecture with approximately 250 sensor nodes along a 25 km double-track mainline and depot facilities, including trackside, rolling-stock, environmental, power, and signaling sensors; data were collected via LoRaWAN and LTE-M gateways and aggregated at local control units and a cloud platform.
- Adapted FMEA with Severity, Occurrence, and Detectability rated 1–10; RPN thresholds were 1–120 acceptable, 121–150 tolerable, and 151–1000 unacceptable; scoring was calibrated through expert elicitation and mapped to CVSS/CWE categories and NIS2 Article 21 Criteria 1 and 6.
- Anomaly scenarios used controlled spoofing, replay with offsets of 100 ms to 5 s, injection with 5–20% amplitude perturbations for vibration data, and timestamp manipulation of 20–80 ms for signaling messages; edge and cloud analytics are described as abstract model classes, and the evaluation is explicitly expert-based and scenario-driven rather than a reproducible ML benchmark.
Limitations
- The paper states limited empirical validation on a single operational segment, data availability constraints due to confidentiality, dynamic threat evolution requiring periodic retraining, and an expert-driven scoring/scenario evaluation that limits reproducibility.
- The paper states it does not include a full experimental protocol expected in ML research, such as formal train/validation/test splits, baseline comparison models, or statistical significance analysis, and omits full ML architectures, hyperparameter grids, resource benchmarks, network-topology diagrams, replay-path instrumentation, and raw log listings.
- The text makes evident that the reported percentages are aggregated scenario-based performance indicators from iterative expert assessments, not outcomes of a reproducible benchmark, and that RPN predictive interpretation requires further empirical validation across broader railway environments.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
12 of 39 cited sources checked · 2 supported · 0 flagged · accuracy not yet measured
- Premisecitation
Most existing research and industrial practice in railway cybersecurity focus on incident response, supply chain security, or high-level governance models, with less attention given to sensor-level risk mechanisms that underpin system reliability.
“However, most existing research and industrial practice in railway cybersecurity focus on incident response, supply chain security, or high-level governance models, with less attention given to the sensor-level risk mechanisms that underpin system reliability [11,12,13].”
Quote in the paperSubstantial claim2 of 3 sources checked- cites [11] Maple, 2017, Security and Privacy in the Internet of Things, J. Cyber Policy source
Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured - cites [12] Ogonji, 2020, A survey on privacy and security of Internet of Things, Comput. Netw. source
- cites [13] Erhan, 2021, Smart Anomaly Detection in Sensor Systems: A Multi-Perspective Review, Inf. Fusion source
Split into 2 atoms — 0 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.
Experimental — accuracy not yet measured
- Premisecitation
The cybersecurity landscape of the railway sector is defined by the coexistence of IT, OT, and sensor networks-each governed by different architectures, standards, and life cycles.
“The cybersecurity landscape of the railway sector is defined by the coexistence of information technology (IT), operational technology (OT), and sensor networks—each governed by different architectures, standards, and life cycles [6,7,15,16,17].”
Quote in the paperSubstantial claim1 of 5 sources checked- cites [6] Wolf, 2018, Safety and Security in Cyber-Physical Systems and Internet-of-Things Systems, Proc. IEEE source
- cites [7] ENISA (2021). Railway Cybersecurity—Good Practices in Cyber Risk Management, European Union Agency for Cybersecurity.
- cites [15] (2018). Failure Modes and Effects Analysis (FMEA and FMECA) (Standard No. IEC 60812:2018).
- cites [16] (2019). Industrial Communication Networks—Network and System Security—System Security Requirements and Security Levels (Standard No. IEC 62443-3-3:2019).
- cites [17] Quan, G., Yao, Z., Si, X., Zhu, W., and Chen, L. (2024). A Data Sharing Model for Blockchain Trusted Sensor Leveraging Mimic Hash Mechanism. Electronics, 13. source
Section 4.2 (Network Deployment) describes a wireless sensor network alongside edge and cloud networks, but does not discuss railway cybersecurity or the coexistence of IT, OT, and sensor networks with different standards and life cycles. Searches for railway IT/OT infrastructure and those distinctions found no passage addressing the claim. (LLM, conf=0.75)
Experimental — accuracy not yet measured
- Premisecitation
A compromised sensor node, falsified measurement, or disrupted data stream can propagate through operational systems, affecting traffic control, rolling stock safety, and signaling integrity.
“A compromised sensor node, falsified measurement, or disrupted data stream can propagate through operational systems, affecting traffic control, rolling stock safety, and signaling integrity [1,2,3,4].”
Quote in the paperSubstantial claim4 of 4 sources checked- cites [1] Ibadah, N., Benavente-Peces, C., and Pahl, M.-O. (2024). Securing the Future of Railway Systems: A Comprehensive Cybersecurity Strategy for Critical On-Board and Track-Side Infrastructure. Sensors, 24. source
- cites [2] Bałdyga, M., Barański, K., Belter, J., Kalinowski, M., and Weichbroth, P. (2024). Anomaly Detection in Railway Sensor Data Environments: State-of-the-Art Methods and Empirical Performance Evaluation. Sensors, 24. source
The Introduction says railway sensors help detect component anomalies and that undetected defects can lead to derailment. Section 5.1 also recognizes missing or incorrect sensor information. Neither passage describes compromised nodes or falsified data propagating through operational systems, or effects on traffic control and signaling integrity. (LLM, conf=0.83)
Experimental — accuracy not yet measured - cites [3] Shi, T., Guo, P., Wang, R., Ma, Z., Zhang, W., Li, W., Fu, H., and Hu, H. (2024). A Survey on Multi-Sensor Fusion Perimeter Intrusion Detection in High-Speed Railways. Sensors, 24. source
The Conclusions (section 7; page unavailable) say that unstable, incomplete, or inaccurate monitoring data sharply reduce intrusion-recognition performance and that better monitoring helps protect trains. The paper does not establish that compromised sensor nodes or disrupted data streams propagate through operational systems, or that they affect traffic control or signaling integrity. (LLM, conf=0.78)
Experimental — accuracy not yet measured - cites [4] Yürekli, Y., Özarpa, C., and Avcı, İ. (2025). Real-Time Railway Hazard Detection Using Distributed Acoustic Sensing and Hybrid Ensemble Learning. Sensors, 25. source
The Introduction discusses detecting environmental hazards with fiber-optic sensors to improve railway safety, and the sensor-setup section describes data collection and transmission. Searches for compromised sensors, falsified measurements, disrupted data streams, and effects on traffic control or signaling found no passage addressing that chain of consequences. (LLM, conf=0.73)
Experimental — accuracy not yet measured
- Interpretationcitation
Sensor networks are critical cybersecurity assets whose reliability directly determines the resilience of railway operations.
“The findings of this study highlight that sensor networks are not only data sources but also critical cybersecurity assets whose reliability directly determines the resilience of railway operations [1,2,7].”
Quote in the paperSubstantial claim2 of 3 sources checked- cites [1] Ibadah, N., Benavente-Peces, C., and Pahl, M.-O. (2024). Securing the Future of Railway Systems: A Comprehensive Cybersecurity Strategy for Critical On-Board and Track-Side Infrastructure. Sensors, 24. source
Section 2.3 describes sensor networks as integral to on-board and trackside operations, says their interaction is critical to safety and efficiency, and identifies their cybersecurity vulnerabilities. The Introduction also calls for making sensors more resilient to cyberattacks. The paper does not establish the stronger claim that sensor-network reliability directly determines the resilience of railway operations; that wording appears to be the citing authors’ inference. (LLM, conf=0.82)
Experimental — accuracy not yet measured - cites [2] Bałdyga, M., Barański, K., Belter, J., Kalinowski, M., and Weichbroth, P. (2024). Anomaly Detection in Railway Sensor Data Environments: State-of-the-Art Methods and Empirical Performance Evaluation. Sensors, 24. source
The Introduction says acoustic sensors are critical for detecting railway-component anomalies that could otherwise lead to derailment. That supports their importance to operational safety, but the paper does not characterize sensor networks as cybersecurity assets or say their reliability directly determines railway resilience. (LLM, conf=0.86)
Experimental — accuracy not yet measured - cites [7] ENISA (2021). Railway Cybersecurity—Good Practices in Cyber Risk Management, European Union Agency for Cybersecurity.
- Findingnone given
In tested scenarios, anomaly detection-detection latency was >60%.
“anomaly detection-detection latency by >60% in tested scenarios”
Quote in the paperNot a substantial claim- Rests on an assumption: The controlled spoofing, replay, and injection anomalies generated in the scenario-based dataset are representative of real cyber-induced sensor failures in railway operations.
- Findingnone given
There was a measurable change in RPN-class distribution (42% Criterion 1/36% Criterion 6).
“measurable change in RPN-class distribution (42% Criterion 1/36% Criterion 6)”
Quote in the paperNot a substantial claim- Rests on an assumption: The controlled spoofing, replay, and injection anomalies generated in the scenario-based dataset are representative of real cyber-induced sensor failures in railway operations.
- Findingreasoning
These results were subsequently prioritized using the calculated Risk Priority Number (RPN) to distinguish the most critical vulnerabilities affecting data integrity, availability, and authenticity.
“These results were subsequently prioritized using the calculated Risk Priority Number (RPN) to distinguish the most critical vulnerabilities affecting data integrity, availability, and authenticity.”
Quote in the paperSubstantial claim - Findingreasoning
These models achieved detection accuracies above 90% for spoofed and replayed signals in simulated tests.
“These models achieved detection accuracies above 90% for spoofed and replayed signals in simulated tests.”
Quote in the paperSubstantial claim - Findingown data
Cross-validation among redundant sensors allowed early detection of spoofed inputs.
“Cross-validation among redundant sensors allowed early detection of spoofed inputs.”
Quote in the paperSubstantial claim - Findingown data
A ±1 change in D leads to an average ±12-18% change in total RPN.
“The analysis demonstrated that a ±1 change in D leads to an average ±12–18% change in total RPN”
Quote in the paperSubstantial claim - Findingreasoning
The research identified a bidirectional dependency between cybersecurity governance and sensor data integrity.
“A key outcome of the research was the identification of a bidirectional dependency between cybersecurity governance and sensor data integrity.”
Quote in the paperSubstantial claim - Findingnone given
A sudden spectral shift unaccompanied by physical cause triggered an alert for potential data injection.
“A sudden spectral shift unaccompanied by physical cause triggered an alert for potential data injection.”
Quote in the paperSubstantial claim - Findingown data
The analysis revealed that most high-priority risks originated from data integrity breaches and limited detectability of anomalies, rather than from hardware malfunctions alone.
“By adapting FMEA to the sensor and IoT context, the analysis revealed that most high-priority risks originated from data integrity breaches and limited detectability of anomalies, rather than from hardware malfunctions alone.”
Quote in the paperSubstantial claim - Findingown data
The FMEA results revealed a strong correlation between detectability (D) and overall risk.
“The FMEA results also revealed a strong correlation between detectability (D) and overall risk,”
Quote in the paperSubstantial claim - Findingown data
In pilot testing, integration of sensor analytics resulted in a 60% decrease in detection latency compared with baseline manual assessments.
“In pilot testing, integration of sensor analytics resulted in a 34% reduction in mean RPN values, a 45% improvement in detection accuracy, and a 60% decrease in detection latency compared with baseline manual assessments.”
Quote in the paperSubstantial claim - Findingown data
The 30 identified cyber failure modes were grouped into four categories corresponding to major threat vectors in sensor networks: Data Integrity Threats, Availability Threats, Configuration and Firmware Threats, and Monitoring and Evaluation Gaps.
“They were grouped into four categories corresponding to major threat vectors in sensor networks: Data Integrity Threats—modification, replay, or spoofing of sensor readings;Availability Threats—communication disruptions, DoS or energy starvation;Configuration and Firmware Threats—unauthorized calibration or malicious firmware updates;Monitoring and Evaluation Gaps—absence of anomaly detection, in”
Quote in the paperNot a substantial claim - Findingnone given
High-risk failure modes are primarily associated with compromised data integrity and insufficient monitoring mechanisms.
“High-risk failure modes are primarily associated with compromised data integrity and insufficient monitoring mechanisms.”
Quote in the paperSubstantial claim - Findingown data
In pilot testing, integration of sensor analytics resulted in a 34% reduction in mean RPN values compared with baseline manual assessments.
“In pilot testing, integration of sensor analytics resulted in a 34% reduction in mean RPN values, a 45% improvement in detection accuracy, and a 60% decrease in detection latency compared with baseline manual assessments.”
Quote in the paperSubstantial claim - Findingreasoning
The integration of redundant verification reduced detection latency from 10 s to 2 s.
“The integration of redundant verification reduced detection latency from 10 s to 2 s”
Quote in the paperSubstantial claim - Findingown data
In this study, anomalies in vibration spectra were correlated with possible data manipulation attempts.
“In this study, anomalies in vibration spectra were also correlated with possible data manipulation attempts.”
Quote in the paperSubstantial claim - Findingreasoning
The adapted FMEA methodology effectively quantified cyber risks associated with sensor networks, providing structured prioritization based on severity, occurrence, and detection parameters.
“The adapted FMEA methodology effectively quantified cyber risks associated with sensor networks, providing structured prioritization based on severity, occurrence, and detection parameters.”
Quote in the paperSubstantial claim - Findingreasoning
Real-time anomaly indicators and streaming sensor evidence were integrated into dynamic RPN updates, enabling measurable risk variation tracking rather than static audit documentation.
“Second, real-time anomaly indicators and streaming sensor evidence were integrated into dynamic RPN updates, enabling measurable risk variation tracking rather than static audit documentation.”
Quote in the paperSubstantial claim - Findingreasoning
The latency reduction from approximately 10 s to 2 s refers to a scenario-based evaluation in which redundant sensor channels (primary + secondary timestamp stream) were used to validate the consistency of event timing.
“The latency reduction from approximately 10 s to 2 s refers to a scenario-based evaluation in which redundant sensor channels (primary + secondary timestamp stream) were used to validate the consistency of event timing.”
Quote in the paperNot a substantial claim - Findingcitation
Based on expert workshops, incident reports, and analysis of (14) and ERA data, 30 cyber failure modes were identified.
“Based on expert workshops, incident reports, and analysis of (14) and ERA data, 30 cyber failure modes were identified.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [14] ENISA (2020). Railway Cybersecurity—Security Measures in the Railway Sector, European Union Agency for Cybersecurity.
- Findingnone given
Detectability (D) is the most influential factor in RPN escalation.
“Detectability (D) is the most influential factor in RPN escalation, emphasizing the importance of sensor data analytics and anomaly detection tools.”
Quote in the paperNot a substantial claim - Findingreasoning
The conversion of sensor-level failure data into RPN metrics provided a transparent basis for prioritizing corrective actions.
“The conversion of sensor-level failure data into RPN metrics provided a transparent basis for prioritizing corrective actions.”
Quote in the paperSubstantial claim - Findingreasoning
Updating RPNs based on live sensor data dynamically quantified how technical and organizational controls influence residual risk, transforming regulatory obligations into measurable, evidence-based indicators.
“By updating Risk Priority Numbers (RPNs) based on live sensor data, the system dynamically quantified how technical and organizational controls influence residual risk, thereby transforming regulatory obligations into measurable, evidence-based indicators.”
Quote in the paperSubstantial claim - Findingown data
In pilot testing, integration of sensor analytics resulted in a 45% improvement in detection accuracy compared with baseline manual assessments.
“In pilot testing, integration of sensor analytics resulted in a 34% reduction in mean RPN values, a 45% improvement in detection accuracy, and a 60% decrease in detection latency compared with baseline manual assessments.”
Quote in the paperSubstantial claim - Findingreasoning
This approach was effective in detecting unusual traffic bursts and irregular vibration sequences indicative of jamming or data injection.
“This approach was effective in detecting unusual traffic bursts and irregular vibration sequences indicative of jamming or data injection.”
Quote in the paperSubstantial claim - Findingreasoning
The integration of quantitative FMEA indicators with real-time monitoring enabled a continuous risk management loop.
“The integration of quantitative FMEA indicators with real-time monitoring enabled a continuous risk management loop.”
Quote in the paperSubstantial claim - Findingown data
The results obtained through the FMEA procedure allowed for the quantitative identification of cyber-induced failure modes at the sensor level.
“The results obtained through the FMEA procedure allowed for the quantitative identification of cyber-induced failure modes at the sensor level.”
Quote in the paperSubstantial claim - Interpretationreasoning
The framework bridges NIS2 regulatory obligations with field-layer monitoring by enabling risk indicators to evolve dynamically rather than remain static documentation artefacts.
“The contribution lies in bridging regulatory obligations of the NIS2 Directive with field-layer monitoring by enabling risk indicators to evolve dynamically rather than remain static documentation artefacts.”
Quote in the paperSubstantial claim - Interpretationreasoning
The sensor-centric FMEA introduced in this study integrates cybersecurity risk factors into the evaluation matrix.
“Unlike traditional FMEA applications, which focus on mechanical or process reliability, the sensor-centric FMEA introduced in this study integrates cybersecurity risk factors into the evaluation matrix.”
Quote in the paperSubstantial claim - Interpretationown data
These metrics demonstrate structural change rather than conceptual evolution.
“These metrics demonstrate structural change rather than conceptual evolution.”
Quote in the paperNot a substantial claim - Study design and conductcitation
The mapping of cybersecurity indicators to FMEA scales was validated through expert elicitation using historical incidents documented by ENISA.
“This mapping was validated through expert elicitation using historical incidents documented by ENISA [7,14].”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [7] ENISA (2021). Railway Cybersecurity—Good Practices in Cyber Risk Management, European Union Agency for Cybersecurity.
- cites [14] ENISA (2020). Railway Cybersecurity—Security Measures in the Railway Sector, European Union Agency for Cybersecurity.
- Interpretationreasoning
The proposed NIS2-FMEA mapping establishes a traceable linkage between regulatory requirements (Article 21), risk categories (impact-likelihood-detectability), and sensor-derived evidence streams.
“The proposed NIS2–FMEA mapping is intentionally process-oriented. Rather than providing a technical audit trail, the model establishes a traceable linkage between regulatory requirements (Article 21), risk categories (impact–likelihood–detectability), and sensor-derived evidence streams.”
Quote in the paperSubstantial claim - Study design and conductreasoning
The scoring scales used in the adapted FMEA model were calibrated using a structured expert-based mapping to established cybersecurity taxonomies, including CVSS v3.1 severity metrics and CWE weakness categories.
“To ensure auditability and methodological transparency, the scoring scales used in the adapted FMEA model were calibrated using a structured expert-based mapping to established cybersecurity taxonomies, including CVSS v3.1 severity metrics and CWE weakness categories.”
Quote in the paperSubstantial claim - Interpretationnone given
The study introduces a sensor-centric cybersecurity framework for railway infrastructure that extends FMEA from traditional reliability evaluation into cyber-induced failures affecting data integrity, availability and authenticity.
“This study introduces a sensor-centric cybersecurity framework for railway infrastructure that extends Failure Mode and Effects Analysis (FMEA) from traditional reliability evaluation into the domain of cyber-induced failures affecting data integrity, availability and authenticity.”
Quote in the paperSubstantial claim - Study design and conductreasoning
Detectability (D) was calibrated through expert assessment of monitoring capabilities, log completeness, timestamp consistency, redundancy levels, and OT/IoT diagnostic mechanisms.
“Detectability (D) was calibrated through expert assessment of monitoring capabilities, log completeness, timestamp consistency, redundancy levels, and OT/IoT diagnostic mechanisms.”
Quote in the paperSubstantial claim - Interpretationreasoning
The central advantage of FMEA lies in its ability to translate qualitative cybersecurity observations into quantifiable indicators-the Risk Priority Number (RPN)-that support compliance with NIS2 Article 21.
“The central advantage of FMEA lies in its ability to translate qualitative cybersecurity observations into quantifiable indicators—the Risk Priority Number (RPN)—that support compliance with NIS2 Article 21.”
Quote in the paperSubstantial claim - Study design and conductnone given
The thresholds used to classify RPN values into acceptable, tolerable, and unacceptable categories were calibrated through three rounds of expert elicitation with specialists in signaling, IoT security, and OT risk assurance.
“The thresholds used to classify RPN values into “acceptable,” “tolerable,” and “unacceptable” categories were calibrated through three rounds of expert elicitation with specialists in signaling, IoT security, and OT risk assurance.”
Quote in the paperSubstantial claim - Study design and conductreasoning
Occurrence (O) incorporated the CVSS Exploitability Sub score and historical CWE frequency patterns.
“Occurrence (O) incorporated the CVSS “Exploitability Sub score” and historical CWE frequency patterns.”
Quote in the paperSubstantial claim - Study design and conductnone given
Each FMEA parameter (S, O, D) is linked to explicit evidence categories: CVSS metrics, CWE categories, sensor monitoring artefacts, log completeness, and redundancy configurations.
“Each FMEA parameter (S, O, D) is linked to explicit evidence categories: CVSS metrics, CWE categories, sensor monitoring artefacts, log completeness, and redundancy configurations.”
Quote in the paperSubstantial claim - Interpretationreasoning
High S and O scores, combined with moderate detectability, indicate systemic policy deficiencies that hinder early identification of threats.
“FMEA linkage: high S and O scores, combined with moderate detectability, indicate systemic policy deficiencies that hinder early identification of threats.”
Quote in the paperSubstantial claim - Interpretationreasoning
By measuring severity, occurrence, and detectability of cyber-related sensor failures, the method produces a prioritized list of vulnerabilities that can be directly mapped to regulatory criteria.
“By measuring the severity (S), occurrence (O), and detectability (D) of cyber-related sensor failures, the method produces a prioritized list of vulnerabilities that can be directly mapped to regulatory criteria.”
Quote in the paperSubstantial claim - Premisereasoning
For the railway sector, NIS2 obligations apply not only to corporate IT systems but also to operational technology (OT) and embedded sensor networks responsible for monitoring tracks, vehicles, and signaling equipment.
“For the railway sector, these obligations apply not only to corporate IT systems but also to operational technology (OT) and embedded sensor networks responsible for monitoring tracks, vehicles, and signaling equipment.”
Quote in the paperSubstantial claim - Premisecitation
The RPN classification ranges (≤120; 121-150; ≥151) reflect harmonization of traditional engineering FMEA interpretations with cybersecurity impact levels inspired by CVSS scoring tiers and NIS2 Article 21(2) treatment requirements.
“The ranges (≤120; 121–150; ≥151) reflect the harmonization of traditional engineering FMEA interpretations [15] with cybersecurity impact levels inspired by CVSS scoring tiers and NIS2 Article 21 (2) treatment requirements.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [15] (2018). Failure Modes and Effects Analysis (FMEA and FMECA) (Standard No. IEC 60812:2018).
- Interpretationreasoning
Railway operators can shift from static compliance verification to dynamic cyber risk detection grounded in real-time evidence.
“Thus, railway operators can shift from static compliance verification to dynamic cyber risk detection grounded in real-time evidence.”
Quote in the paperSubstantial claim - Interpretationreasoning
Cybersecurity and data governance cannot be separated in cyber-physical systems such as railway networks.
“This interdependence implies that cybersecurity and data governance cannot be separated in cyber–physical systems such as railway networks.”
Quote in the paperSubstantial claim - Interpretationnone given
The work extended the traditional FMEA model beyond hardware and reliability failures, incorporating cyber-induced failure modes affecting data integrity, availability, and authenticity in railway IoT and OT environments.
“First, the work extended the traditional Failure Mode and Effects Analysis (FMEA) model beyond hardware and reliability failures, incorporating cyber-induced failure modes affecting data integrity, availability, and authenticity in railway IoT and OT environments.”
Quote in the paperSubstantial claim - Interpretationreasoning
By integrating sensor data into cybersecurity dashboards, operators can detect deviations that signal potential cyber incidents.
“By integrating sensor data into cybersecurity dashboards, operators can detect deviations that signal potential cyber incidents.”
Quote in the paperSubstantial claim - Interpretationreasoning
The quality and completeness of sensor data directly affect the ability to verify policy effectiveness and compliance with NIS2.
“On the other hand, the quality and completeness of sensor data directly affect the ability to verify policy effectiveness and compliance with NIS2.”
Quote in the paperSubstantial claim - Interpretationreasoning
The mechanism relies on parallel timestamp comparison rather than on a specific network topology or buffering architecture.
“The mechanism relies on parallel timestamp comparison rather than on a specific network topology or buffering architecture.”
Quote in the paperNot a substantial claim - Interpretationnone given
Sensor networks constitute the foundation of cybersecurity in cyber-physical railway systems.
“Sensor networks constitute the foundation of cybersecurity in cyber–physical railway systems.”
Quote in the paperNot a substantial claim - Study design and conductnone given
Autoencoders and clustering algorithms (e.g., DBSCAN) were deployed to discover outliers in multi-sensor data streams without requiring prior labeling.
“Autoencoders and clustering algorithms (e.g., DBSCAN) were deployed to discover outliers in multi-sensor data streams without requiring prior labeling.”
Quote in the paperSubstantial claim - Study design and conductreasoning
Thresholds for anomaly alerts were adapted using simple percentile-based or median absolute deviation heuristics, aligned with expert consensus rather than formal optimization.
“Thresholds for anomaly alerts were adapted using simple percentile-based or median absolute deviation heuristics, aligned with expert consensus rather than formal optimization.”
Quote in the paperSubstantial claim - Interpretationreasoning
Extending FMEA to encompass cyber-induced sensor failures demonstrated how regulatory obligations can be operationalized through quantitative, data-driven indicators.
“By extending the Failure Mode and Effects Analysis (FMEA) approach to encompass cyber-induced sensor failures, the research demonstrated how regulatory obligations can be operationalized through quantitative, data-driven indicators.”
Quote in the paperSubstantial claim - Study design and conductreasoning
The before and after latency values reflect expert-observed differences in anomaly-notification delay when redundancy checks are enabled, rather than the outcome of a fully instrumented network-performance benchmark.
“The “before” and “after” latency values reflect expert-observed differences in anomaly-notification delay when redundancy checks are enabled, rather than the outcome of a fully instrumented network-performance benchmark.”
Quote in the paperNot a substantial claim - Study design and conductreasoning
The reported percentages should be interpreted as aggregated scenario-based performance indicators derived from iterative expert assessments, rather than as outcomes of a reproducible benchmark with formal training/validation/test splits.
“Accordingly, the reported percentages should be interpreted as aggregated scenario-based performance indicators derived from iterative expert assessments, rather than as outcomes of a reproducible benchmark with formal training/validation/test splits.”
Quote in the paperSubstantial claim - Interpretationown data
Detectability is the dominant factor in cyber-induced sensor failures.
“confirming that detectability is the dominant factor in cyber-induced sensor failures.”
Quote in the paperSubstantial claim - Study design and conductreasoning
The expert-driven scoring model and scenario-based anomaly evaluation limited reproducibility due to the absence of large-scale labelled datasets and inter-operator benchmarking.
“The expert-driven scoring model and scenario-based anomaly evaluation limited reproducibility, particularly due to the absence of large-scale labelled datasets and inter-operator benchmarking.”
Quote in the paperSubstantial claim - Study design and conductreasoning
The reported quantitative indicators (detection accuracy improved by 45%, detection latency reduced by 60%, and mean RPN reduced by 34%) do not originate from a single, fully reproducible machine-learning experiment based on a closed dataset.
“It is important to clarify that the reported quantitative indicators (“detection accuracy improved by 45%, detection latency reduced by 60%, and mean RPN reduced by 34%”) do not originate from a single, fully reproducible machine-learning experiment based on a closed dataset.”
Quote in the paperSubstantial claim - Interpretationnone given
The proposed framework provides a feedback mechanism between field-level sensors and strategic governance processes.
“It provides a feedback mechanism between field-level sensors and strategic governance processes.”
Quote in the paperSubstantial claim - Interpretationreasoning
The incorporation of these data into Security Information and Event Management (SIEM) or Operational Technology (OT) Security Operation Center (SOC) platforms enables cross-domain correlation between physical and digital indicators.
“The incorporation of these data into Security Information and Event Management (SIEM) or Operational Technology (OT) Security Operation Center (SOC) platforms enables cross-domain correlation between physical and digital indicators.”
Quote in the paperNot a substantial claim - Interpretationown data
These improvements demonstrate the tangible benefits of embedding sensor analytics into the cybersecurity management cycle mandated by NIS2.
“These improvements demonstrate the tangible benefits of embedding sensor analytics into the cybersecurity management cycle mandated by NIS2.”
Quote in the paperSubstantial claim - Interpretationreasoning
The study provided a structured methodology that operationalized NIS2 risk-management requirements through sensor-derived indicators and FMEA quantification.
“This study provided a structured methodology that operationalized NIS2 risk-management requirements through sensor-derived indicators and FMEA quantification.”
Quote in the paperSubstantial claim - Interpretationnone given
The proposed FMEA-sensor integration framework supports compliance harmonization between NIS2 and sector-specific standards such as EN 50126/50129 (RAMS) and ISO/IEC 27001.
“It supports compliance harmonization between NIS2 and sector-specific standards such as EN 50126 [20]/50129 [21] (RAMS) and ISO/IEC 27001 [18].”
Quote in the paperSubstantial claim0 of 3 sources checked- cites [20] (2017). Railway Applications—The Specification and Demonstration of Reliability, Availability, Maintainability and Safety (RAMS)—Part 1: Generic RAMS Process (Standard No. EN 50126-1:2017).
- cites [21] (2018). Railway Applications—Communication, Signalling and Processing Systems—Safety-Related Electronic Systems for Signalling (Standard No. EN 50129:2018).
- cites [18] (2013). Information Security Management Systems—Requirements (Standard No. ISO/IEC 27001:2013).
- Interpretationown data
Poor visibility into sensor data integrity substantially amplifies cybersecurity exposure.
“indicating that poor visibility into sensor data integrity substantially amplifies cybersecurity exposure.”
Quote in the paperSubstantial claim - Interpretationreasoning
Cyber-induced sensor failures are a core category of operational hazards in modern railway infrastructure.
“Therefore, risk analysis in modern railway infrastructure must account for cyber-induced sensor failures as a core category of operational hazards.”
Quote in the paperSubstantial claim - Interpretationnone given
Deviations from baseline patterns in traction power sensors can indicate malicious firmware tampering affecting measurement calibration.
“Deviations from baseline patterns can indicate not only equipment overloads but also malicious firmware tampering affecting measurement calibration.”
Quote in the paperSubstantial claim - Interpretationreasoning
The vulnerability of a vibration or axle counter sensor to spoofing or firmware manipulation can have consequences equivalent to a physical fault in safety-critical systems.
“In practice, the vulnerability of a vibration or axle counter sensor to spoofing or firmware manipulation can have consequences equivalent to a physical fault in safety-critical systems.”
Quote in the paperSubstantial claim - Interpretationreasoning
The approach provided verifiable metrics (e.g., RPN, MTTD, DA, FAR) that can serve as objective evidence in cybersecurity audits and performance reviews mandated by competent authorities under NIS2.
“The approach provided verifiable metrics (e.g., RPN, MTTD, DA, FAR) that can serve as objective evidence in cybersecurity audits and performance reviews mandated by competent authorities under NIS2.”
Quote in the paperSubstantial claim - Study design and conductnone given
Classification algorithms such as Random Forest or Support Vector Machines were trained on labeled datasets comprising normal and compromised sensor states.
“Classification algorithms such as Random Forest or Support Vector Machines were trained on labeled datasets comprising normal and compromised sensor states.”
Quote in the paperSubstantial claim - Premisecitation
Insufficient monitoring and incomplete logs are major amplifiers of systemic cyber risk.
“This aligns with industry observations and ENISA guidance, which indicate that insufficient monitoring and incomplete logs are major amplifiers of systemic cyber risk.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites ENISA guidance
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationreasoning
ML algorithms, trained on historical operational data, can identify subtle deviations that may precede failures or cyber intrusions.
“ML algorithms, trained on historical operational data, can identify subtle deviations that may precede failures or cyber intrusions.”
Quote in the paperSubstantial claim - Interpretationreasoning
The FMEA model facilitated collaboration between safety engineers and cybersecurity specialists.
“The FMEA model facilitated collaboration between safety engineers and cybersecurity specialists.”
Quote in the paperNot a substantial claim - Interpretationown data
High RPN values for spoofing and firmware tampering indicate insufficient governance over sensor lifecycle management and patch control.
“High RPN values for spoofing and firmware tampering indicate insufficient governance over sensor lifecycle management and patch control.”
Quote in the paperSubstantial claim - Study design and conductreasoning
The reported quantitative indicators result from an expert-based evaluation approach and comparative assessment of several operational scenarios conducted in a demonstration environment.
“Instead, they result from an expert-based evaluation approach (expert elicitation) and comparative assessment of several operational scenarios conducted in a demonstration environment.”
Quote in the paperSubstantial claim - Interpretationown data
Compliance with NIS2 Article 21 in the railway sector cannot be achieved without addressing cybersecurity at the sensor and data layer.
“The case study findings confirm that compliance with NIS2 Article 21 in the railway sector cannot be achieved without addressing cybersecurity at the sensor and data layer.”
Quote in the paperSubstantial claim - Study design and conductreasoning
Lightweight models were implemented directly on edge gateways to minimize response times.
“To minimize response times, lightweight models were implemented directly on edge gateways.”
Quote in the paperSubstantial claim - Interpretationreasoning
The integration of FMEA with live monitoring enables continuous reassessment of risks, directly supporting the NIS2 requirement for periodic evaluation of control effectiveness.
“The integration of FMEA with live monitoring enables continuous reassessment of risks, directly supporting the NIS2 requirement for periodic evaluation of control effectiveness.”
Quote in the paperSubstantial claim - Interpretationown data
Organizational and monitoring deficiencies-rather than purely technical failures-represent the dominant contributors to high-risk scores.
“This confirms that organizational and monitoring deficiencies—rather than purely technical failures—represent the dominant contributors to high-risk scores.”
Quote in the paperSubstantial claim - Interpretationreasoning
Well-defined policies covering configuration management, authentication, and monitoring enable early detection of anomalies in sensor networks.
“On one hand, well-defined policies—covering configuration management, authentication, and monitoring—enable early detection of anomalies in sensor networks.”
Quote in the paperSubstantial claim - Interpretationreasoning
The contributions filled a methodological gap between governance frameworks and field-layer sensor analytics.
“Together, these contributions filled a methodological gap between governance frameworks and field-layer sensor analytics.”
Quote in the paperNot a substantial claim - Interpretationown data
Sensor-level cybersecurity plays a pivotal role in ensuring the resilience of railway operations.
“The case study demonstrates that sensor-level cybersecurity plays a pivotal role in ensuring the resilience of railway operations.”
Quote in the paperSubstantial claim - Study design and conductnone given
The case study network included trackside, rolling-stock and environmental measurement devices, reflecting sampling and communication characteristics typically observed in operational railway systems.
“The network included trackside, rolling-stock and environmental measurement devices, reflecting sampling and communication characteristics typically observed in operational railway systems.”
Quote in the paperSubstantial claim- Rests on an assumption: The studied railway IoT architecture is representative of typical operational railway systems.
- Interpretationreasoning
Effective policy enforcement requires sensor-generated evidence supporting continuous evaluation in addition to procedural documentation.
“Effective policy enforcement requires not only procedural documentation but also sensor-generated evidence supporting continuous evaluation.”
Quote in the paperSubstantial claim - Interpretationreasoning
Continuous recalibration of RPN values based on streaming sensor data enabled the transition from static, compliance-driven audits to predictive cybersecurity.
“Continuous recalibration of RPN values based on streaming sensor data enabled the transition from static, compliance-driven audits to predictive cybersecurity, in which risks are anticipated and mitigated before incidents occur.”
Quote in the paperSubstantial claim - Study design and conductnone given
Labels for normal and anomalous conditions were generated through expert interpretation of recorded operating data and supplemented with controlled spoofing, replay and data-injection scenarios.
“Labels for “normal” and “anomalous” conditions were generated through expert interpretation of recorded operating data and supplemented with controlled spoofing, replay and data-injection scenarios.”
Quote in the paperSubstantial claim - Interpretationcitation
The same principles apply to safety-critical transportation systems.
“The current work extends these insights to the railway sector, showing that the same principles apply to safety-critical transportation systems.”
Quote in the paperNot a substantial claim - Interpretationreasoning
Failures linked to ineffective anomaly detection and missing audit trails demonstrate the need for continuous monitoring systems that integrate sensor data analytics into cybersecurity dashboards.
“Failures linked to ineffective anomaly detection and missing audit trails demonstrate the need for continuous monitoring systems that integrate sensor data analytics into cybersecurity dashboards.”
Quote in the paperSubstantial claim - Interpretationown data
Cyber resilience is a measurable property that can be continuously observed through sensor-derived metrics.
“From a systems perspective, the study demonstrates that cyber resilience is a measurable property that can be continuously observed through sensor-derived metrics.”
Quote in the paperSubstantial claim
- Probably not a substantial claim. 86
- It may state two separate claims as one. 45
- The quote only partly states this claim. 1
Probably not a substantial claim. The paper’s conclusions may not depend on it.
PremisecitationCyber threats affecting railway sensors and IoT systems can be grouped into three main categories: data integrity attacks, availability disruptions, and confidentiality breaches.
“Cyber threats affecting railway sensors and IoT systems can be grouped into three main categories [1,5,14]:Data integrity attacks, such as spoofing, injection, or tampering with measurement values.”
- Also stated in: § 2. Regulatory and Technological Context · plain
“Availability disruptions, including denial-of-service (DoS) attacks or network congestion, which can delay sensor communication or disable critical monitoring functions.”
- Also stated in: § 2. Regulatory and Technological Context · hedged
“Confidentiality breaches, where unauthorized access to sensor data may reveal sensitive operational parameters or be used for targeted attacks on rolling stock and signaling subsystems.”
Quote in the paperNot a substantial claim1 of 3 sources checked- cites [1] Ibadah, N., Benavente-Peces, C., and Pahl, M.-O. (2024). Securing the Future of Railway Systems: A Comprehensive Cybersecurity Strategy for Critical On-Board and Track-Side Infrastructure. Sensors, 24. source
Section 3.2 (page not provided) names confidentiality, integrity, and availability as railway security objectives, and Section 2.3 describes attacks that could manipulate or corrupt sensor data. The paper does not explicitly group threats to railway sensors and IoT systems into those three attack categories; the cited passages also do not give spoofing, injection, and measurement tampering as a set of examples. (LLM, conf=0.84)
Experimental — accuracy not yet measured - cites [5] Fernandes, 2025, Cybersecurity in Smart Railways: Exploring risks, vulnerabilities and mitigation in the data communication services, Glob. Energy Interconnect. Trans. source
- cites [14] ENISA (2020). Railway Cybersecurity—Security Measures in the Railway Sector, European Union Agency for Cybersecurity.
- Also stated in: § 2. Regulatory and Technological Context · plain
It may state two separate claims as one.
PremisecitationStudies published in Sensors have shown that anomaly detection, blockchain-based integrity verification, and AI-driven analytics enhance trust in industrial sensor networks.
“Studies published in Sensors (e.g., [19,22,23,24]) have shown that anomaly detection, blockchain-based integrity verification, and AI-driven analytics enhance trust in industrial sensor networks.”
Quote in the paperNot a substantial claim2 of 4 sources checked- cites [19] (2019). Security and Resilience—Business Continuity Management Systems—Requirements (Standard No. ISO 22301:2019).
- cites [22] Govea, J., Gaibor-Naranjo, W., and Villegas-Ch, W. (2024). Securing Critical Infrastructure with Blockchain Technology: An Approach to Cyber-Resilience. Computers, 13. source
- cites [23] Aslam, A., Postolache, O., Oliveira, S., and Pereira, J.D. (2025). Securing IoT Sensors Using Sharding-Based Blockchain Network Technology Integration: A Systematic Review. Sensors, 25. source
The Introduction describes blockchain records as tamper-resistant and protective of transaction integrity. Section 3.1 discusses anomaly detection in IoT systems and the potential of AI and blockchain for intrusion-detection systems. The paper does not establish that all three approaches enhance trust specifically in industrial sensor networks; its discussion of trustworthiness in Section 4.2 concerns blockchain-based IoT sensor security. (LLM, conf=0.84)
Experimental — accuracy not yet measured - cites [24] Saleh, O.A., and Cevik, M. (2025). Blockchain-Integrated Secure Authentication Framework for Smart Grid IoT Using Energy-Aware Consensus Mechanisms. Sensors, 25. source
Probably not a substantial claim.
FindingreasoningFailure modes such as absence of asset inventory, undefined responsibilities for IoT security, or lack of firmware update policy fall into the Criterion 1 group.
“Failure modes such as absence of asset inventory, undefined responsibilities for IoT security, or lack of firmware update policy fall into this group.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataEvidence from real sensor streams can recalibrate O and D factors in near-real-time and reduce undetected exposure windows, enabling measurable compliance documentation aligned with NIS2 Article 21.
“Instead of presenting a fixed benchmark, the results show how evidence from real sensor streams can recalibrate O and D factors in near-real-time and reduce undetected exposure windows, enabling measurable compliance documentation aligned with NIS2 Article 21.”
Quote in the paperSubstantial claim- Rests on an assumption: The controlled spoofing, replay, and injection anomalies generated in the scenario-based dataset are representative of real cyber-induced sensor failures in railway operations.
- Rests on an assumption: The approximately 250 trackside, rolling-stock, environmental and power-monitoring sensors deployed over a 25 km operational segment are representative of railway infrastructure more broadly.
It may state two separate claims as one.
Findingown dataApproximately 42% of identified risks were related to Criterion 1 (risk analysis and information system security), 36% to Criterion 6 (control effectiveness assessment), and the remaining 22% to shared or cross-domain issues.
“Figure 2 illustrates this distribution, showing that approximately 42% of identified risks were related to Criterion 1 (risk analysis and information system security), 36% to Criterion 6 (control effectiveness assessment), and the remaining 22% to shared or cross-domain issues.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
FindingreasoningIntegration of real-time sensor monitoring enabled dynamic updates of RPN values, establishing a continuous feedback loop between risk assessment and control effectiveness evaluation, directly addressing NIS2 Criteria 1 and 6.
“Integration of real-time sensor monitoring enabled dynamic updates of RPN values, establishing a continuous feedback loop between risk assessment and control effectiveness evaluation—directly addressing NIS2 Criteria 1 and 6.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
FindingreasoningLower, yet still significant risks included API vulnerabilities and environmental sensor tampering, primarily due to moderate detectability and lack of redundancy in local data validation.
“Lower, yet still significant risks included API vulnerabilities and environmental sensor tampering, primarily due to moderate detectability and lack of redundancy in local data validation.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
FindingreasoningBased on the observed reduction in detectability (D) and occurrence (O), the RPN score is recalculated and reclassified from tolerable to unacceptable.
“Based on the observed reduction in detectability (D) and occurrence (O), the RPN score is recalculated and reclassified from “tolerable” to “unacceptable.””
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataThe FMEA-based analysis revealed two key dependency loops relevant to regulatory implementation: Policy-Technology Alignment (Criterion 1) and Monitoring-Evaluation Loop (Criterion 6).
“The FMEA-based analysis revealed two key dependency loops relevant to regulatory implementation: Policy–Technology Alignment (Criterion 1) ○High RPN values for spoofing and firmware tampering indicate insufficient governance over sensor lifecycle management and patch control.○NIS2 compliance therefore requires explicit inclusion of IoT/OT asset inventories, firmware assurance processes, and config”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingnone givenMachine learning and edge analytics significantly improved detectability, reducing detection latency and enabling predictive risk management.
“Machine learning and edge analytics significantly improved detectability, reducing detection latency and enabling predictive risk management.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Findingown dataSecure firmware validation through checksum monitoring was deployed, enabling automated risk reassessment with a reduction in RPN from 405 to 270.
“Secure firmware validation through checksum monitoring was deployed, enabling automated risk reassessment (reduction in RPN from 405 to 270).”
Quote in the paperSubstantial claimProbably not a substantial claim.
FindingreasoningIn an axle-counter replay scenario, the edge-layer model flags an anomaly due to timestamp inconsistency.
“In an axle-counter replay scenario, a previously recorded sequence of occupancy messages is re-injected with a temporal offset. The edge-layer model flags an anomaly due to timestamp inconsistency”
Quote in the paperSubstantial claimIt may state two separate claims as one.
FindingreasoningThese models perform local inference, flagging abnormal sensor behavior before forwarding alerts to the central platform.
“These models perform local inference, flagging abnormal sensor behavior before forwarding alerts to the central platform.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
Findingnone givenRedundancy checks consisted of comparing sequential timestamp streams from two independent sensing paths, allowing anomalies caused by replay offsets to be detected earlier without requiring changes to the underlying communication network.
“Redundancy checks consisted of comparing sequential timestamp streams from two independent sensing paths, allowing anomalies caused by replay offsets to be detected earlier without requiring changes to the underlying communication network.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Findingown dataThe mean RPN value across all analyzed modes was 326, with the top quartile exceeding 430, a level classified as unacceptable under the adopted risk scale.
“The mean RPN value across all analyzed modes was 326, with the top quartile exceeding 430, a level classified as unacceptable under the adopted risk scale.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisereasoningSupply chain exploitation can introduce hidden backdoors through compromised firmware or counterfeit components.
“Supply Chain Exploitation—As sensors are sourced from multiple vendors, compromised firmware or counterfeit components may introduce hidden backdoors.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
Study design and conductreasoningSeverity (S) was aligned with the CVSS Impact Sub score.
“Severity (S) was aligned with the CVSS “Impact Sub score,””
Quote in the paperSubstantial claimProbably not a substantial claim.
InterpretationreasoningAmong NIS2's ten categories of cybersecurity measures, Criterion 1 (risk analysis and information system security policy) and Criterion 6 (policies and procedures for assessing effectiveness of risk management measures) are particularly relevant for sensor-based railway systems.
“While the NIS2 Directive outlines ten categories of cybersecurity measures, two are particularly relevant for sensor-based railway systems:Criterion 1: Risk analysis and information system security policy, andCriterion 6: Policies and procedures for assessing the effectiveness of risk management measures.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may be too vague to check against evidence.
InterpretationreasoningThere is a need for structured analytical tools that can transform abstract compliance requirements into quantifiable risk indicators directly associated with sensor network performance and integrity.
“As a result, there is a need for structured analytical tools that can transform abstract compliance requirements into quantifiable risk indicators directly associated with sensor network performance and integrity.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
PremisecitationThe NIS2 Directive defines new obligations for operators of essential services, including railway transport.
“Within the European Union, the NIS2 Directive (Directive (EU) 2022/2555) defines new obligations for operators of essential services, including railway transport [5,6,7,8,9,10].”
Quote in the paperSubstantial claim0 of 6 sources checked- cites [5] Fernandes, 2025, Cybersecurity in Smart Railways: Exploring risks, vulnerabilities and mitigation in the data communication services, Glob. Energy Interconnect. Trans. source
- cites [6] Wolf, 2018, Safety and Security in Cyber-Physical Systems and Internet-of-Things Systems, Proc. IEEE source
- cites [7] ENISA (2021). Railway Cybersecurity—Good Practices in Cyber Risk Management, European Union Agency for Cybersecurity.
- cites [8] (2025, November 02). Directive (EU) 2022/2555 of the European Parliament and of the Council of 14 December 2022 on Measures for a High Common Level of Cybersecurity across the Union (NIS2 Directive). Official Journal of the European Union, L 333. Available online: https://eur-lex.europa.eu/eli/dir/2022/2555/oj/eng.
- cites [9] Ruohonen, J. (2024). A Systematic Literature Review on the NIS2 Directive. arXiv.
- cites [10] Castiglione, G., Santamaria, D.F., and Bella, G. (2023). NIS2Onto: An Ontological Approach to Compliance Verification of the NIS2 Directive. arXiv.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
InterpretationreasoningCriterion 1 and Criterion 6 jointly form a continuous improvement loop.
“These two provisions jointly form a continuous improvement loop.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Premisenone givenThe ranges for mapping to detectability reflect monitoring maturity assessments commonly used in NIS2 conformity audits.
“These ranges reflect the monitoring maturity assessments commonly used in NIS2 conformity audits.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
InterpretationreasoningThe chain of evidence is conceptual but verifiable at the procedural level: each FMEA dimension corresponds to a documented NIS2 obligation, and each is supported by measurable artefacts such as sensor logs, anomaly alerts, or detection timestamps.
“The chain of evidence is therefore conceptual but verifiable at the procedural level: each FMEA dimension corresponds to a documented NIS2 obligation (e.g., monitoring, anomaly detection, incident handling), and each is supported by measurable artefacts, such as sensor logs, anomaly alerts, or detection timestamps, which can be referenced during compliance assessments.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
PremisereasoningCriterion 6 relates to the evaluation of controls applied to mitigate sensor-related threats.
“Criterion 6—Effectiveness Assessment of Risk Management MeasuresRelates to the evaluation of controls applied to mitigate sensor-related threats.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Study design and conductreasoningThe calibration was intentionally conservative: any scenario with O or D ≥ 8 automatically pushes the RPN into the unacceptable tier, even with S in the medium range, reflecting the high regulatory importance of detectability under NIS2.
“The calibration was intentionally conservative: any scenario with O or D ≥ 8 automatically pushes the RPN into the “unacceptable” tier, even with S in the medium range, reflecting the high regulatory importance of detectability under NIS2.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
InterpretationreasoningIn sensor networks, the cybersecurity of each sensing element, its communication interface, and its data-handling processes must be verifiable and traceable.
“In the context of sensor networks, this means that the cybersecurity of each sensing element, its communication interface, and its data-handling processes must be verifiable and traceable.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds.
InterpretationreasoningFMEA can provide a bridge between compliance mandates and technical evaluation by quantifying the likelihood, severity, and detectability of cyber-related sensor failures.
“Tools such as the Failure Mode and Effects Analysis (FMEA) can provide a bridge between compliance mandates and technical evaluation by quantifying the likelihood, severity, and detectability of cyber-related sensor failures.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
InterpretationreasoningRisk identification through structured approaches like FMEA provides a quantitative basis for prioritizing mitigation measures and integrating them into an organization's cybersecurity policy (Criterion 1).
“Furthermore, risk identification through structured approaches like FMEA provides a quantitative basis for prioritizing mitigation measures and integrating them into an organization’s cybersecurity policy (Criterion 1).”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
PremisereasoningUnder NIS2, operators must demonstrate evidence-based compliance by documenting risk assessments, performance indicators (KPIs/KRIs), and audit trails.
“Moreover, under NIS2, operators must demonstrate evidence-based compliance—for example, by documenting risk assessments, performance indicators (KPIs/KRIs), and audit trails.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds.
InterpretationreasoningReal-time analytics can identify abnormal patterns in sensor behavior, signal inconsistencies, or timing anomalies that indicate potential compromise.
“Real-time analytics can identify abnormal patterns in sensor behavior, signal inconsistencies, or timing anomalies that indicate potential compromise.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds. It may state two separate claims as one.
PremisereasoningSensor data manipulation attacks can mislead algorithms responsible for predictive maintenance or condition monitoring; falsified axle counter data or incorrect vibration readings could trigger erroneous control actions.
“Sensor Data Manipulation Attacks (Integrity)—Adversaries may alter measurement values, replay valid data packets, or inject false readings. Such attacks can mislead algorithms responsible for predictive maintenance or condition monitoring. In railway contexts, falsified axle counter data or incorrect vibration readings could trigger erroneous control actions.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
InterpretationreasoningUnder NIS2, monitoring directly supports Criterion 6 by enabling operators to demonstrate the functionality and reliability of cybersecurity controls.
“Under the NIS2 framework, such monitoring directly supports Criterion 6—assessment of effectiveness, by enabling operators to demonstrate the functionality and reliability of cybersecurity controls.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
InterpretationreasoningThis layered structure ensures operational efficiency but also propagates cascading vulnerabilities-a compromise at the edge or communication layer can propagate upward, affecting decision-making at supervisory or cloud levels.
“This layered structure ensures operational efficiency but also propagates cascading vulnerabilities—a compromise at the edge or communication layer can propagate upward, affecting decision-making at supervisory or cloud levels.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
PremisereasoningUnauthorized access and malware infiltration can allow adversaries to use compromised sensor gateways as entry points into OT networks.
“Unauthorized Access and Malware Infiltration (Confidentiality and Control)—Weak authentication or outdated firmware can allow unauthorized access to sensor gateways. Once compromised, adversaries can use these devices as entry points into OT networks.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
InterpretationreasoningThese challenges directly relate to the NIS2 obligation of continuous improvement.
“These challenges directly relate to the NIS2 obligation of continuous improvement.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
InterpretationreasoningThe linkage of FMEA parameters to evidence categories creates a clear evidence chain for conformity assessments performed by national authorities, ensuring that the risk scoring process is transparent and repeatable even if the underlying datasets are not publicly released.
“This creates a clear evidence chain for conformity assessments performed by national authorities, ensuring that the risk scoring process is transparent and repeatable even if the underlying datasets are not publicly released.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper may give no reason why it holds.
InterpretationreasoningThe NIS2 Directive implicitly encompasses these risks through its emphasis on technical and organizational measures for ensuring system security.
“The NIS2 Directive implicitly encompasses these risks through its emphasis on technical and organizational measures for ensuring system security.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may be too vague to check against evidence.
InterpretationreasoningTranslating regulatory requirements into technical implementations for sensor systems requires a structured risk assessment methodology.
“However, translating regulatory requirements into technical implementations for sensor systems requires a structured risk assessment methodology.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may state two separate claims as one.
Study design and conductreasoningThe calibration steps, mapping tables, and justification of thresholds were documented in a structured expert assessment protocol and are therefore auditable according to NIS2 governance requirements.
“The calibration steps, mapping tables, and justification of thresholds were documented in a structured expert assessment protocol and are therefore auditable according to NIS2 governance requirements.”
Quote in the paperSubstantial claimProbably not a substantial claim.
InterpretationreasoningFulfilling NIS2 obligations requires methods capable of linking technical vulnerabilities of sensors and IoT systems with organizational risk governance and compliance monitoring.
“In practice, fulfilling these obligations requires methods capable of linking technical vulnerabilities of sensors and IoT systems with organizational risk governance and compliance monitoring.”
Quote in the paperSubstantial claimProbably not a substantial claim.
Premisenone givenCriterion 1 corresponds to risks arising from inadequate identification, documentation, or governance of vulnerabilities in sensor networks.
“Criterion 1—Risk Analysis and Security PolicyCorresponds to risks arising from inadequate identification, documentation, or governance of vulnerabilities in sensor networks.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may be too vague to check against evidence. The paper may give no reason why it holds.
InterpretationreasoningTechnological convergence between IT, OT, and IoT requires a sensor-centric cybersecurity architecture that combines device-level protection, encrypted communication, and centralized monitoring of integrity and performance.
“Technological convergence between IT, OT, and IoT therefore requires a sensor-centric cybersecurity architecture—one that combines device-level protection, encrypted communication, and centralized monitoring of integrity and performance.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
PremisereasoningDenial-of-service and jamming can prevent timely transmission of sensor data; even short interruptions can lead to operational delays or degraded safety margins.
“Denial-of-Service and Jamming (Availability)—Attackers may overload communication channels or jam radio frequencies, preventing timely transmission of sensor data. In critical systems, even short interruptions can lead to operational delays or degraded safety margins.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds. It may state two separate claims as one.
InterpretationreasoningIn this architecture, FMEA-based risk assessment serves as a foundation for identifying critical failure modes in sensor systems and for aligning mitigation priorities with the regulatory requirements of NIS2.
“In this architecture, FMEA-based risk assessment serves as a foundation for identifying critical failure modes in sensor systems and for aligning mitigation priorities with the regulatory requirements of NIS2.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
PremisereasoningThe interconnection of IT, OT, and IoT components in railway systems introduces a broad and evolving cyber threat landscape, which can be categorized according to the primary target and intended impact.
“The interconnection of IT, OT, and IoT components in railway systems introduces a broad and evolving cyber threat landscape, which can be categorized according to the primary target and intended impact:”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may state two separate claims as one.
PremisereasoningThis hybrid architecture introduces multiple integration challenges: heterogeneity of communication protocols, legacy interoperability, limited computational resources at the edge, data synchronization and validation, and security monitoring and anomaly detection.
“This hybrid architecture introduces multiple integration challenges:Heterogeneity of communication protocols—sensors may use protocols such as M-Bus, Modbus, or MQTT, which vary in encryption capabilities and authentication mechanisms.”
Quote in the paperSubstantial claimProbably not a substantial claim.
PremisecitationSeverity is tied to operational safety impact and safety margin reduction defined in EN 50126/50129 risk categories.
“Severity is tied to operational safety impact and safety margin reduction defined in EN 50126/50129 [20,21] risk categories.”
Quote in the paperSubstantial claim0 of 2 sources checked- cites [20] (2017). Railway Applications—The Specification and Demonstration of Reliability, Availability, Maintainability and Safety (RAMS)—Part 1: Generic RAMS Process (Standard No. EN 50126-1:2017).
- cites [21] (2018). Railway Applications—Communication, Signalling and Processing Systems—Safety-Related Electronic Systems for Signalling (Standard No. EN 50129:2018).
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Interpretationnone givenThe architecture of railway sensor systems can be conceptualized as a four-tier model comprising edge, communication, control, and cloud layers.
“The architecture of railway sensor systems can be conceptualized as a four-tier model comprising edge, communication, control, and cloud layers (Figure 1).”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Premisenone givenAggregated sensor data can inadvertently reveal sensitive operational patterns.
“Data Correlation and Privacy Risks—Aggregated sensor data can inadvertently reveal sensitive operational patterns.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds. It may state two separate claims as one.
Interpretationnone givenEach of the elements used to calibrate the scoring scales is routinely used in compliance assessments performed by national competent authorities under NIS2, allowing the scoring model to be auditable and reproducible at the procedural level.
“Each of these elements is routinely used in compliance assessments performed by national competent authorities under NIS2, allowing the scoring model to be auditable and reproducible at the procedural level.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds.
Interpretationnone givenThe same data that serve maintenance and safety purposes can also be harnessed to enhance cybersecurity monitoring and threat detection.
“The same data that serve maintenance and safety purposes can also be harnessed to enhance cybersecurity monitoring and threat detection.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Interpretationnone givenLoss of synchronization in timestamps may point to compromised gateways or tampered communication modules.
“Loss of synchronization in timestamps may point to compromised gateways or tampered communication modules.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
InterpretationreasoningAbnormal latency in data transmission may indicate network congestion or a denial-of-service attack.
“Abnormal latency in data transmission may indicate network congestion or a denial-of-service attack.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
InterpretationreasoningThe integrated framework supports a transition from reactive compliance to proactive cyber resilience, where the railway system continuously validates its own security posture through sensor-derived intelligence.
“This integrated framework thus supports a transition from reactive compliance to proactive cyber resilience, where the railway system continuously validates its own security posture through sensor-derived intelligence.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may be too vague to check against evidence.
Interpretationnone givenCybersecurity risk management in critical infrastructure must evolve from document-based frameworks toward evidence-based, data-centric systems.
“The conducted research has shown that cybersecurity risk management in critical infrastructure must evolve from document-based frameworks toward evidence-based, data-centric systems.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
Study design and conductreasoningGiven the expert-driven and scenario-oriented character of the study, the edge and cloud analytical components represent abstracted model classes rather than specific, fully parameterized architectures.
“Given the expert-driven and scenario-oriented character of the study, the edge and cloud analytical components represent abstracted model classes rather than specific, fully parameterized architectures.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it.
InterpretationreasoningCyber-threat evolution requires periodic recalibration of the Occurrence (O) and Detection (D) parameters, as fixed expert estimates may lose validity over time without continuous evidence ingestion.
“Cyber-threat evolution required periodic recalibration of the Occurrence (O) and Detection (D) parameters, as fixed expert estimates may lose validity over time without continuous evidence ingestion.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may state two separate claims as one.
Study design and conductreasoningThe methodology follows an expert-based evaluation paradigm and does not include the full experimental protocol typically expected in ML research.
“Given the focus of the study—linking NIS2 governance processes with sensor-driven analytical workflows—the methodology follows an expert-based evaluation paradigm. It does not include the full experimental protocol typically expected in ML research (e.g., formalized train/validation/test splits, baseline comparison models, or statistical significance analysis).”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningThe integration of real-time sensor analytics into risk management supports technical resilience and provides documented evidence for supervisory authorities.
“The integration of real-time sensor analytics into risk management not only supports technical resilience but also provides documented evidence for supervisory authorities.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningThe TCMS event was mapped to NIS2 Criterion 6, confirming the effectiveness of continuous anomaly evaluation.
“The event was mapped to NIS2 Criterion 6, confirming the effectiveness of continuous anomaly evaluation.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
InterpretationreasoningA sudden increase in communication errors may elevate the Occurrence (O) parameter.
“a sudden increase in communication errors may elevate the Occurrence (O) parameter;”
Quote in the paperNot a substantial claimProbably not a substantial claim.
Study design and conductnone givenThe detection performance values presented in Section 6 should be understood as synthesized results from a set of controlled test scenarios combined with expert cybersecurity assessments, rather than statistically validated outcomes of a fully reproducible ML pipeline.
“The detection performance values presented in Section 6 should be understood as synthesized results from a set of controlled test scenarios combined with expert cybersecurity assessments, rather than statistically validated outcomes of a fully reproducible ML pipeline.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may state two separate claims as one.
Study design and conductreasoningThese exercises were intended to create representative disturbance patterns to support expert evaluation of FMEA parameters, rather than to constitute formal penetration-testing campaigns or adversarial ML benchmarks.
“These exercises were intended to create representative disturbance patterns to support expert evaluation of FMEA parameters, rather than to constitute formal penetration-testing campaigns or adversarial ML benchmarks.”
Quote in the paperSubstantial claimProbably not a substantial claim.
Study design and conductreasoningThese models are representative abstractions whose purpose is to illustrate analytical workflow and its influence on FMEA parameters, rather than to optimize machine-learning performance.
“These models are representative abstractions whose purpose is to illustrate analytical workflow and its influence on FMEA parameters, rather than to optimize machine-learning performance.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Study design and conductown dataThe case study relied on a single operational segment, so broader validation across multiple networks is needed to ensure generalizability.
“Limited empirical validation: The case study relied on a single operational segment; broader validation across multiple networks is needed to ensure generalizability.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may be too vague to check against evidence.
InterpretationreasoningThere is growing convergence between safety engineering and cybersecurity, as both domains rely on the accuracy and availability of measurement data.
“This insight confirms the growing convergence between safety engineering and cybersecurity, as both domains rely on the accuracy and availability of measurement data.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
InterpretationreasoningFailure modes such as missing performance indicators, ineffective anomaly detection algorithms, or absence of post-incident review reflect weaknesses in the feedback loop required by NIS2.
“Failure modes such as missing performance indicators, ineffective anomaly detection algorithms, or absence of post-incident review reflect weaknesses in the feedback loop required by NIS2.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningThe FMEA outputs were explicitly mapped onto regulatory obligations defined in NIS2 Article 21, demonstrating that compliance could be evidenced through data-driven KPIs such as detection latency, MTTD, and post-mitigation RPN contraction.
“Third, the FMEA outputs were explicitly mapped onto regulatory obligations defined in NIS2 Article 21, demonstrating that compliance could be evidenced through data-driven KPIs such as detection latency, Mean Time to Detect (MTTD), and post-mitigation RPN contraction.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningThis continuous risk management loop satisfies the dual intent of NIS2 Article 21: establishing structured risk analysis procedures (Criterion 1) and continuously evaluating the effectiveness of implemented measures (Criterion 6).
“This loop satisfies the dual intent of NIS2 Article 21: (1) establishing structured risk analysis procedures (Criterion 1) and (2) continuously evaluating the effectiveness of implemented measures (Criterion 6).”
Quote in the paperSubstantial claimProbably not a substantial claim. It may be too vague to check against evidence. It may state two separate claims as one.
Interpretationown dataRailway cybersecurity cannot be confined to network protection or compliance documentation; it must extend to the sensor layer, where data originate and where operational resilience is determined in real time.
“Overall, the study demonstrated that railway cybersecurity cannot be confined to network protection or compliance documentation—it must extend to the sensor layer, where data originate and where operational resilience is determined in real time.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may state two separate claims as one.
InterpretationreasoningThis integration creates a closed-loop system where sensor data not only support operational decisions but also inform cyber risk management and regulatory compliance tracking.
“This integration creates a closed-loop system where sensor data not only support operational decisions but also inform cyber risk management and regulatory compliance tracking.”
Quote in the paperSubstantial claimProbably not a substantial claim.
InterpretationreasoningSuch decentralization enhances system resilience by reducing dependence on cloud connectivity.
“Such decentralization enhances system resilience by reducing dependence on cloud connectivity.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. The paper may give no reason why it holds. It may state two separate claims as one.
Interpretationnone givenWhile traditional rule-based monitoring can detect predefined anomalies, the complexity of sensor environments in railways requires machine learning (ML) and edge analytics for adaptive detection of emerging threats.
“While traditional rule-based monitoring can detect predefined anomalies, the complexity of sensor environments in railways requires machine learning (ML) and edge analytics for adaptive detection of emerging threats.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds.
InterpretationreasoningThe proposed FMEA model can be integrated with real-time monitoring systems that process sensor data streams from IoT and OT devices.
“To ensure scalability and automation, the proposed FMEA model can be integrated with real-time monitoring systems that process sensor data streams from IoT and OT devices.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductown dataThe demonstration environment incorporated typical resource constraints observed in railway IoT/edge deployments.
“The demonstration environment also incorporated typical resource constraints observed in railway IoT/edge deployments.”
Quote in the paperNot a substantial claim- Rests on an assumption: The demonstration environment's resource constraints are typical of railway IoT/edge deployments.
Probably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
InterpretationreasoningA recurring fault type raises the overall RPN priority.
“and a recurring fault type raises the overall RPN priority.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds.
InterpretationreasoningContinuous monitoring provides evidence of control functionality, enabling railway operators to demonstrate Criterion 6 compliance through measurable, sensor-derived KPIs.
“By providing continuous evidence of control functionality, such monitoring enables railway operators to demonstrate Criterion 6 compliance through measurable, sensor-derived KPIs.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may be too vague to check against evidence.
InterpretationreasoningThe proposed approach offers a practical mechanism for aligning technical risk assessment with governance-level compliance processes.
“The proposed approach offers a practical mechanism for aligning technical risk assessment with governance-level compliance processes.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
InterpretationreasoningInconsistent measurement patterns across redundant sensors may suggest data spoofing or unauthorized calibration.
“Inconsistent measurement patterns across redundant sensors may suggest data spoofing or unauthorized calibration.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Interpretationnone givenAs attack vectors evolve, periodic retraining of detection algorithms is required to maintain effectiveness.
“Dynamic threat evolution: As attack vectors evolve, periodic retraining of detection algorithms is required to maintain effectiveness.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
InterpretationreasoningThe approach provides a transparent, evidence-based method for demonstrating compliance to national competent authorities and supervisory bodies.
“Finally, the approach provides a transparent, evidence-based method for demonstrating compliance to national competent authorities and supervisory bodies.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningThe absence of telemetry from specific nodes prevents verification of incident response efficiency or system availability, undermining indicators required for demonstrating regulatory conformity.
“For example, the absence of telemetry from specific nodes prevents verification of incident response efficiency or system availability, thus undermining the very indicators required for demonstrating regulatory conformity.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper may give no reason why it holds.
InterpretationreasoningContinuous analysis of anomalies, deviations, or missing data packets provides feedback to update the FMEA parameters dynamically.
“Continuous analysis of anomalies, deviations, or missing data packets provides feedback to update the FMEA parameters dynamically.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
Study design and conductreasoningHyperparameters were selected through expert tuning aimed at illustrating behavioral differences rather than optimizing ML performance.
“Hyperparameters were selected through expert tuning aimed at illustrating behavioral differences rather than optimizing ML performance.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
InterpretationreasoningThe study proposed and validated a sensor-centric framework for cybersecurity risk management in railway infrastructure, aligning the methodology with NIS2 requirements.
“This study proposed and validated a sensor-centric framework for cybersecurity risk management in railway infrastructure, aligning the methodology with the requirements of the NIS2 Directive.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Interpretationnone givenDelayed detection of an anomaly increases Detection (D).
“delayed detection of an anomaly increases Detection (D);”
Quote in the paperNot a substantial claimProbably not a substantial claim.
Study design and conductnone givenCyberattack scenarios including spoofing, replay, data injection, and communication jamming were implemented using lightweight scripts designed to approximate typical anomalies rather than to emulate full adversarial toolchains.
“Cyberattack scenarios—including spoofing, replay, data injection, and communication jamming—were implemented using lightweight scripts designed to approximate typical anomalies rather than to emulate full adversarial toolchains.”
Quote in the paperSubstantial claim- Rests on an assumption: The lightweight scripts used for cyberattack scenarios approximate typical anomalies in operational railway sensor networks.
Probably not a substantial claim. The paper may give no reason why it holds.
Interpretationnone givenThe proposed framework offers a scalable methodology applicable to both legacy OT assets and emerging IoT platforms.
“It offers a scalable methodology applicable to both legacy OT assets and emerging IoT platforms.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Study design and conductnone givenThe qualitative sensitivity analysis is fully auditable and follows a structured expert-evaluation process.
“Although not a formal statistical sensitivity study, the analysis is fully auditable and follows a structured expert-evaluation process.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductreasoningThe case-study network description does not include full implementation details such as exact ML architectures, hyperparameter grids, or resource benchmarks due to the scenario-based and expert-driven nature of the study.
“Due to the scenario-based and expert-driven nature of the study, the case-study network description does not include full implementation details such as exact ML architectures, hyperparameter grids, or resource benchmarks.”
Quote in the paperNot a substantial claimIt may state two separate claims as one.
Interpretationnone givenTheir correct functioning and data integrity are essential for ensuring both operational safety and compliance with NIS2.
“Their correct functioning and data integrity are essential for ensuring both operational safety and compliance with NIS2.”
Quote in the paperSubstantial claimProbably not a substantial claim. It may state two separate claims as one.
Study design and conductreasoningThe sample sizes were intentionally limited-the study was not designed as a full-scale big-data experiment-and individual scenario windows ranged from approximately 30 min to 2 h.
“The sample sizes were intentionally limited—the study was not designed as a full-scale big-data experiment—and individual scenario windows ranged from approximately 30 min to 2 h.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Interpretationown dataSensor data analytics directly enhance cybersecurity performance indicators, bridging the gap between technical operations and governance-level reporting.
“These examples confirm that sensor data analytics directly enhance cybersecurity performance indicators, bridging the gap between technical operations and governance-level reporting.”
Quote in the paperSubstantial claimIt may state two separate claims as one.
Study design and conductreasoningThe evaluation combines three sources of evidence: recorded sensor streams from selected railway infrastructure devices, simulated spoofing and replay events executed in a controlled test setting, and structured assessments provided by cybersecurity experts responsible for IoT and signaling systems.
“The evaluation combines three sources of evidence: (i) recorded sensor streams from selected railway infrastructure devices, (ii) simulated spoofing and replay events executed in a controlled test setting, and (iii) structured assessments provided by cybersecurity experts responsible for IoT and signaling systems.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper may give no reason why it holds.
Premisenone givenIn the context of NIS2 compliance, continuous monitoring is essential to demonstrate that risk management measures remain effective (Criterion 6).
“In the context of NIS2 compliance, continuous monitoring is essential to demonstrate that risk management measures remain effective (Criterion 6).”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductnone givenThe demonstration scenarios employed a set of abstract model classes representing analytical approaches commonly used in OT/IoT monitoring systems.
“The demonstration scenarios employed a set of abstract model classes representing analytical approaches commonly used in OT/IoT monitoring systems.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Interpretationnone givenIntegration of edge analytics into legacy OT systems demands careful performance balancing, especially where safety certification cycles restrict frequent software changes.
“Furthermore, integration of edge analytics into legacy OT systems demanded careful performance balancing, especially where safety certification cycles restrict frequent software changes.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may be too vague to check against evidence. The paper may give no reason why it holds. It may state two separate claims as one.
InterpretationreasoningPositioning sensors as active components of the cybersecurity architecture allows the railway sector to meet regulatory obligations and enhance the safety, reliability, and trustworthiness of its operations.
“By positioning sensors as active components of the cybersecurity architecture, the railway sector can not only meet regulatory obligations but also enhance the safety, reliability, and trustworthiness of its operations.”
Quote in the paperNot a substantial claimProbably not a substantial claim.
InterpretationreasoningThe resulting artefacts-detection logs, timestamps, anomaly scores, and updated RPN classification-constitute the audit-relevant evidence supporting NIS2 Article 21 requirements.
“The resulting artefacts—detection logs, timestamps, anomaly scores, and updated RPN classification—constitute the audit-relevant evidence supporting NIS2 Article 21 requirements.”
Quote in the paperSubstantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
InterpretationreasoningThe approach contributes to the development of digital twins for infrastructure cybersecurity.
“Furthermore, the approach contributes to the development of digital twins for infrastructure cybersecurity—virtual models that continuously assimilate sensor data to simulate and evaluate the impact of potential cyber events.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductnone givenThe analysis used a combination of real and synthetically generated data, including vibration, temperature, axle-load, and selected signaling telemetry parameters.
“The analysis used a combination of real and synthetically generated data, including vibration, temperature, axle-load, and selected signaling telemetry parameters.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Interpretationnone givenBy continuously adapting detection thresholds through learning algorithms, the system can dynamically adjust FMEA parameters (Occurrence and Detection), enabling data-driven updates of the risk profile.
“By continuously adapting detection thresholds through learning algorithms, the system can dynamically adjust FMEA parameters (Occurrence and Detection), enabling data-driven updates of the risk profile.”
Quote in the paperNot a substantial claimIt may state two separate claims as one. The quote only partly states this claim.
InterpretationreasoningThe reduction in detection latency from 10 s to 2 s significantly improved system resilience and audit traceability under NIS2 requirements.
“significantly improving system resilience and audit traceability under NIS2 requirements.”
Quote in the paperSubstantial claimProbably not a substantial claim.
PremisecitationUnder Article 21 of NIS2, essential entities must establish procedures for testing, auditing, and evaluating the effectiveness of controls.
“Under Article 21 of NIS2, essential entities must establish procedures for testing, auditing, and evaluating the effectiveness of controls.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites Article 21 of NIS2
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
Study design and conductreasoningThere is a methodological gap in linking RPN variation with resilience forecasting; RPN reduction demonstrates control effectiveness retrospectively, but its predictive interpretation requires further empirical validation.
“There also remained a methodological gap in linking RPN variation with resilience forecasting: while RPN reduction demonstrated control effectiveness retrospectively, its predictive interpretation required further empirical validation across broader railway environments.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper may give no reason why it holds.
InterpretationreasoningQuantified sensor-based indicators, including detection rate and MTTD, can serve as verifiable evidence in compliance audits conducted by national supervisory authorities, fulfilling NIS2 Article 32 requirements.
“Quantified sensor-based indicators, including detection rate and Mean Time to Detect (MTTD), can serve as verifiable evidence in compliance audits conducted by national supervisory authorities, fulfilling NIS2 Article 32 requirements regarding oversight and reporting.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it. The paper may give no reason why it holds.
Study design and conductnone givenFeature selection followed an expert-based approach consistent with predictive maintenance practices in railway infrastructure.
“Feature selection followed an expert-based approach consistent with predictive maintenance practices in railway infrastructure.”
Quote in the paperNot a substantial claimProbably not a substantial claim. The paper’s conclusions may not depend on it.
Study design and conductnone givenSome security-relevant datasets remain restricted due to confidentiality, limiting the full application of supervised ML models.
“Data availability constraints: Some security-relevant datasets remain restricted due to confidentiality, limiting the full application of supervised ML models.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may be too vague to check against evidence.
InterpretationreasoningSensor-enabled cybersecurity governance represents a crucial step toward the digital resilience envisioned by the NIS2 Directive and the broader European strategy for secure, intelligent transport systems.
“In conclusion, sensor-enabled cybersecurity governance—supported by structured methodologies such as FMEA and empowered by real-time analytics—represents a crucial step toward the digital resilience envisioned by the NIS2 Directive and the broader European strategy for secure, intelligent transport systems.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may be too vague to check against evidence.
InterpretationreasoningData quality management is a component of cybersecurity governance frameworks.
“Consequently, data quality management should be considered a component of cybersecurity governance frameworks, with explicit inclusion in risk management documentation submitted to national authorities under NIS2 supervision.”
Quote in the paperNot a substantial claimProbably not a substantial claim. It may state two separate claims as one.
InterpretationreasoningHigh D values (poor detectability) suggest insufficient monitoring and lack of continuous evaluation mechanisms.
“FMEA linkage: high D values (poor detectability) suggest insufficient monitoring and lack of continuous evaluation mechanisms.”
Quote in the paperSubstantial claim
The paper relies on this without support.
The controlled spoofing, replay, and injection anomalies generated in the scenario-based dataset are representative of real cyber-induced sensor failures in railway operations.
“with representative anomalies generated through controlled spoofing, replay and injection conditions”
The paper relies on this without support.
The approximately 250 trackside, rolling-stock, environmental and power-monitoring sensors deployed over a 25 km operational segment are representative of railway infrastructure more broadly.
“approximately 250 trackside, rolling-stock, environmental and power-monitoring sensors deployed over a 25 km operational segment”
The paper relies on this without support.
The studied railway IoT architecture is representative of typical operational railway systems.
“The case study was conducted on a representative railway IoT architecture comprising approximately 250 sensor nodes deployed along a 25 km double-track mainline and associated depot facilities.”
“reflecting sampling and communication characteristics typically observed in operational railway systems.”
The paper relies on this without support.
The lightweight scripts used for cyberattack scenarios approximate typical anomalies in operational railway sensor networks.
“Cyberattack scenarios—including spoofing, replay, data injection, and communication jamming—were implemented using lightweight scripts designed to approximate typical anomalies rather than to emulate full adversarial toolchains.”
The paper relies on this without support.
The demonstration environment's resource constraints are typical of railway IoT/edge deployments.
“The demonstration environment also incorporated typical resource constraints observed in railway IoT/edge deployments.”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Decreasing both O and D cannot raise the RPN category
MajorThe example says reductions in O and D reclassify the risk from tolerable to unacceptable. With severity unchanged, reducing both factors can only decrease their product and therefore cannot raise the RPN category.
“Based on the observed reduction in detectability (D) and occurrence (O), the RPN score is recalculated and reclassified from “tolerable” to “unacceptable.””
“The Risk Priority Number (RPN) is calculated as:RPN=S×O×D where each parameter is rated on a 1–10 scale.”
An O or D score of 8 does not automatically make risk unacceptable
MajorThe stated RPN formula and 1-10 scales allow a medium severity of 5, occurrence of 8, and detectability score of 1 to produce an RPN of 40-not an unacceptable score of at least 151. Thus an O score of 8 cannot automatically put every such scenario in the unacceptable tier.
“any scenario with O or D ≥ 8 automatically pushes the RPN into the “unacceptable” tier, even with S in the medium range”
“The Risk Priority Number (RPN) is calculated as:RPN=S×O×D where each parameter is rated on a 1–10 scale.”
Citations
Cites
Smart Anomaly Detection in Sensor Systems: A Multi-Perspective Review
Erhan, L., Ndubuaku, M., Di Mauro, M., Song, W., Chen, M., Fortino, G., Bagdasar, O., & Liotta, A. (2021) · Information Fusion · doi:10.1016/j.inffus.2020.10.001 · Cited by 378
Security and Privacy in the Internet of Things
Maple, C. (2017) · Journal of Cyber Policy · doi:10.1080/23738871.2017.1366536 · Cited by 316
Safety and Security in Cyber-Physical Systems and Internet-of-Things Systems
Wolf, M., & Serpanos, D. (2018) · Proceedings of the IEEE · doi:10.1109/jproc.2017.2781198 · Cited by 250
A survey on privacy and security of Internet of Things
Ogonji, M. M., Okeyo, G., & Wafula, J. M. (2020) · Computer Science Review · doi:10.1016/j.cosrev.2020.100312 · Cited by 230
A Survey on Multi-Sensor Fusion Perimeter Intrusion Detection in High-Speed Railways
Shi, T., Guo, P., Wang, R., Ma, Z., Zhang, W., Li, W., Fu, H., & Hu, H. (2024) · Sensors · doi:10.3390/s24175463 · Cited by 42
Bałdyga, M., Barański, K., Belter, J., Kalinowski, M., & Weichbroth, P. (2024) · Sensors · doi:10.3390/s24082633 · Cited by 40
Ibadah, N., Benavente-Peces, C., & Pahl, M.-O. (2024) · Sensors · doi:10.3390/s24248218 · Cited by 39
A Systematic Literature Review on the NIS2 Directive
Ruohonen, J. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2412.08084 · Cited by 37
Securing Critical Infrastructure with Blockchain Technology: An Approach to Cyber-Resilience
Govea, J., Gaibor-Naranjo, W., & Villegas-Ch, W. (2024) · Computers · doi:10.3390/computers13050122 · Cited by 33
Fernandes, T., Magalhães, J. P., & Alves, W. (2025) · Green Energy and Intelligent Transportation · doi:10.1016/j.geits.2025.100305 · Cited by 23
Real-Time Railway Hazard Detection Using Distributed Acoustic Sensing and Hybrid Ensemble Learning
Yürekli, Y., Özarpa, C., & Avcı, İ. (2025) · Sensors · doi:10.3390/s25133992 · Cited by 21
Saleh, O. A., & Cevik, M. (2025) · Sensors · doi:10.3390/s25216622 · Cited by 18
Aslam, A., Postolache, O., Oliveira, S., & Pereira, J. D. (2025) · Sensors · doi:10.3390/s25030807 · Cited by 10
A Data Sharing Model for Blockchain Trusted Sensor Leveraging Mimic Hash Mechanism
Quan, G., Yao, Z., Si, X., Zhu, W., & Chen, L. (2024) · Electronics · doi:10.3390/electronics13081495 · Cited by 9
Flagged first, then by citations
Research works printed but not linked: 2
- (2025, November 02). Directive (EU) 2022/2555 of the European Parliament and of the Council of 14 December 2022 on Measures for a High Common Level of Cybersecurity across the Union (NIS2 Directive). Official Journal of the European Union, L 333. Available online: https://eur-lex.europa.eu/eli/dir/2022/2555/oj/eng.
- Castiglione, G., Santamaria, D.F., and Bella, G. (2023). NIS2Onto: An Ontological Approach to Compliance Verification of the NIS2 Directive. arXiv.
Cited by
Sorted by citations
Cognitive Load and Compliance: A Human-Centric Framework for NIS2 in Latvian SMEs
Breidaks, I., Kaļķis, H., & Семенов, А. (2026) · AHFE international · doi:10.54941/ahfe1007408 · Cited by 1
Cyber-Physical Railway Security using Smart Port Operations
Shrivastav, R. (2026) · Journal of Intelligent Decision Making and Information Science · doi:10.59543/jidmis.v3.663
Cybersecurity Aspects of Artificial Intelligence Driven System for Railway Infrastructure Telemetry
Atanasov, I. I., & Pencheva, E. N. (2026) · IEEE International Workshop on Metrology for Industry 4.0 & IoT (MetroInd4.0&IoT) · doi:10.1109/metroind4.0iot69397.2026.11653108
Marpaung, Y. E., & Lany, A. (2026) · Jurnal Iman dan Spiritualitas · doi:10.15575/jis.v5i4.58614
Metadata
- Authors
- Wachnik, R., Chruzik, K., & Pochopień, B.
- Affiliations
- WSB University (PL)
- Corresponding author
- Wachnik, Rafał
- Venue
- Sensors
- Publisher
- MDPI AG
- Published
- 2025-12-04
- Volume
- 25
- Issue
- 23
- Pages
- 7384
- ISSN
- 1424-8220 (ISSN-L)
- PMC
- PMC12694181
- OpenAlex
- W4417273874
- Semantic Scholar
- 2c28fb66f0d33afef41dfd0ce5679107b7d51cc1
- Cited by
- 4
- References
- 11
- Citations by year
- 2026: 4
- FWCI
- 2.41
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Sensors (publishedVersion, cc-by)
- PubMed (publishedVersion)
- PubMed Central (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Physical Sciences › Computer Science › Information Systems › Information and Cyber Security
- Physical Sciences › Engineering › Control and Systems Engineering › Transportation Systems and Safety
- Physical Sciences › Engineering › Control and Systems Engineering › Smart Grid Security and Resilience
- Keywords
- NIS2 Directive, failure mode and effects analysis, risk priority number, anomaly detection, data integrity, regulatory compliance, cybersecurity governance
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-11-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
