Research Briefs on Information and Communication Technology Evolution

Evaluation of an Integrated Disaster Response Support System Utilizing Information and Communication Technology

Yuta Seri — 2026-02-05

In this study, we developed an integrated disaster response support system to improve the efficiency of disaster response operations and integrate information management in local governments. In response to recent issues, e.g., a shortage of specialized staff and difficulties with knowledge transfer, the proposed system was designed to integrate functions, support disaster operations using the latest technologies, and eliminate the dependency on personal knowledge. The results of an effectiveness evaluation by local government employees demonstrated the system’s potential to eliminate information fragmentation and support decision-making processes. In addition, a usability evaluation using the system usability scale rated it highly, confirming that even inexperienced staff can operate the proposed system intuitively. Furthermore, a chatbot equipped with retrieval-augmented generation functionality demonstrated high accuracy rates and efficiency in several information search tasks. The findings of this study indicate that the proposed system combines ease of use with advanced technology, thereby making it a useful platform that will contribute to improving the disaster response capabilities of local governments.

SGX-Enabled Encrypted Storage for Secure Management of 5G Authentication Data in Trusted Execution Environments

Adi Panca Saputra Iskandar — 2026-02-05

The fifth-generation (5G) core network, sensitive authentication data such as the Subscription Permanent Identifier (SUPI) and long-term cryptographic keys are centrally stored in the Unified Data Repository (UDR) to support primary authentication. Recent real-world incidents, including the 2025 SK Telecom (SKT) breach, demonstrate that compromise of core servers or databases can expose plaintext subscriber data even when transport-layer security is correctly deployed. This high- lights the need for strong data-at-rest protection and robust cryptographic key isolation within the 5G core. In this paper, we propose a storage-centric protection scheme that preserves the confidentiality and integrity of 5G authentication data even under authentication server compromise. Authentication records are encrypted before being stored in the UDR, with encryption and decryption operations in- voked through database triggers and executed inside an Intel Software Guard Extensions (SGX)-based Trusted Execution Environment (TEE). All cryptographic keys and sensitive operations are fully iso- lated within the enclave, preventing direct access from both the database and application layers. We implement the proposed design on the OpenAirInterface (OAI) 5G core using a MySQL-backed UDR, demonstrating its applicability to real-world and open-source 5G deployments. Performance evalua- tion over 10,000 end to end authentication procedures shows that the proposed approach introduces moderate CPU overhead, particularly during decryption-intensive operations, while incurring negli- gible memory overhead and minimal latency impact. These results indicate that SGX-based storage centric protection is a practical and effective mechanism for strengthening data at rest security in 5G core networks.

Benchmarking Deep Learning Architectures for Real-Time Intrusion Detection in Kubernetes-Orchestrated 5G Core Networks

Vincent Abella — 2026-02-05

Cloud-native 5G core networks on Service-Based Architecture expose distributed Network Functions to cyber threats requiring adaptive Deep Learning-based Intrusion Detection Systems (DL-IDS). This work evaluates six DL architectures (Convolutional Neural Network (CNN), Multi-Layer Perceptron (MLP), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Autoencoder (AE)) on a Kubernetes-orchestrated Open5GS testbed, measuring Central Processing Unit (CPU) utilization, memory consumption, and latency under realistic traffic conditions. Results show feedforward models (CNN, MLP, AE) achieve sub-millisecond latency (0.6 milliseconds (ms)) with CPU below 12%, enabling multiple concurrent IDS instances per server, while recurrent architectures (RNN, LSTM, GRU) require high CPU utilization (99-107%) with 3.5 to 7.2 ms latency, necessitating dedicated hardware acceleration. Memory footprint remains consistent (385 to 390 megabytes (MB)) across all models. These findings demonstrate that operational efficiency is a key consideration for DL-IDS deployment in production 5G networks, with substantial CPU efficiency differences between architecture choices.

A Lifecycle-Based Security Threat Model for FPGA in Safety-Critical Systems

Dongmin Kim — 2026-03-01

Field-programmable gate arrays are increasingly adopted in safety-critical systems due to their deterministic execution, low latency, and suitability for rigorous verification and validation. However, prior studies largely focus on individual attack techniques or the operational phase, limiting their ability to capture how security threats are introduced, propagate, and remain dormant across the FPGA development lifecycle. To address this limitation, this paper proposes a lifecycle-based security threat analysis framework that integrates the IEEE Std 1012 verification and validation lifecycle with the FPGA development flow. The proposed framework supports systematic analysis by anchoring the assessment to key development artifacts spanning from design through bitstream generation and operation. By structuring FPGA security threats from a lifecycle perspective, this study provides a foundational analytical framework for systematically analyzing security risks across the FPGA development lifecycle in safety-critical systems.

Visual Analysis of Outdoor Surveillance Videos Using Principal Component Analysis

Kaoru Sugita — 2026-03-06

In this paper, we investigated the application of principal component analysis (PCA) to the visualization of outdoor videos for safety and security monitoring. We analyzed videos depicting daytime airplane takeoffs and landings, nighttime airplane landings, small birds flying, and small birds resting on an elevated bridge. The extracted frames were arranged on a two-dimensional plane according to their principal component scores. The results indicate that frame placement reflects inter-frame correlations and is strongly influenced by global visual factors such as illumination conditions and the size of moving subjects. A qualitative evaluation suggests that the visualization provides an intuitive overview of frame-level variations and reduces the workload required for scene exploration without continuous playback. However, the effectiveness decreases for scenes involving small or low-contrast subjects. These findings clarify the characteristics and limitations of PCA-based visualization for outdoor surveillance videos.

A BERT-Based Question Answering System for Optical Fiber Transmission Networks

Haiyang Wang — 2026-06-03

To help operations staff quickly get technical support and decision-making basis for highway fiber-optic transmission networks, this paper plans to develop a BERT-based Question Answering (QA) system. The system first builds a knowledge graph to store related knowledge. Then, it uses the BERT model to identify entities and intents in user questions, matches them in the knowledge graph via template matching, and returns answers. This system offers efficient information query and technical support for highway fiber-optic transmission networks. Experimental results show that BERT model has 10.01%, 6.21%, 3.72% higher F1-score than TextRNN, TextCNN, Fasttext in intent recognition. BERT-BiLSTM-CRF model has 10.01%, 6.21%, 3.72% higher F1-score than LSTM, BiLSTM, BiLSTM-CRF in entity recognition.

A Novel Power Data Sharing Platform

Zhang Siwen — 2026-09-18

The expansion of advanced metering, distributed energy resources, dispatch operations, asset management, and electricity-market data has exposed the limitations of traditional database-to-database exchange and manually administered authorization. Such conventional solutions cannot simultaneously achieve semantic interoperability, cross-domain least-privilege access, low-latency delivery, privacy minimization, and end-to-end accountability. This paper presents CECF, the Cloud-Edge-Chain Feedback platform for power-data sharing. The IEC Common Information Model provides the semantic foundation. Edge gateways perform tasks such as protocol adaptation, time-stamp correction, data quality filtering, sensitive field masking, stream data buffering and hotspot content caching. The cloud data plane integrates object storage services, time-series databases and event bus components. The authorization framework combines role-based access control, attribute-based security policies, access purpose declaration and real-time context risk evaluation. Large operational payloads remain encrypted off-chain. Only message digests, policy versions, timestamps, authorization decisions and audit evidence indexes are recorded on a permissioned ledger. Using monitoring indicators including P95 latency, queue length, cache-hit rate, failure rate and policy risk value, the feedback controller adjusts evidence micro-batch parameters, cache prefetching strategies and policy thresholds. A differential-privacy interface supports statistical data sharing scenarios. The reported simulations use Poisson arrival flow and multi-stage queueing models. At an offered load of 1200 requests per second, CECF achieves P95 latency of 16.38 ms. Relative to centralized architecture, full on-chain scheme and static-hybrid architecture, its P95 latency drops by 42.1%, 69.5% and 14.6% respectively. When the offered request rate rises to 1800 requests/s, the platform maintains 1802.8 SLO-qualified requests per second. The ablation results indicate that edge caching and closed-loop feedback control are the principal mechanisms that prevent a sharp increase in tail latency. The design provides a feasible route to power-data sharing with high efficiency, verifiability and standardized governance. Its parameters nevertheless require recalibration on a testbed close to production conditions.

Visualizing Post-Election Presidential Approval Ratings in Texas and California: A County-Level Geospatial Analysis Using Eight Complementary Visualization Techniques

Sultan Fahim — 2026-09-18

While polarization has been extensively studied on the national level, there is notable geographical variability within states and counties when it comes to presidential approval ratings. In this paper, we investigate county-level temporal and geographical variations of presidential approval ratings in two contrasting states, namely Republican Texas and Democratic California, from January 2020 to October 2025. Based on aggregated presidential approval rating estimates provided by Gallup, Pew Research Center, and FiveThirtyEight, and merged using county-level shapefiles obtained from the U.S. Census Bureau TIGER/Line with the help of FIPS codes, we create an analysis workflow that results in eight different visualizations. Our major contribution is the justification for each visualization type based on Munzner's task decomposition method, showing how each visualized statistic answers its own research question not addressable by any other type. While California consistently registers approval ratings above the national average by 8-12 percentage points, Texas falls behind by 6-10 percentage points; the difference in approval ratings increased by approximately 5 points following the 2024 presidential elections. A continuous positive association between approval ratings and an urban-rural axis exists in both states (r = +0.71 in California and r = +0.58 in Texas). Finally, a population-weighted bubble map proves that national disapproval visible in choropleths is a consequence of geography rather than demography. We claim that visualization design choices in political science are epistemological: what gets shown in a chart affects the nature of analysis.

AI Large Model-Based Data Value Mining in Power Projects

Zhang Jialin — 2026-09-25

Power project usually has the characteristics of long construction period, many participants, complex supply chain and large number of project documents. Information such as schedule, earned value, cost, contract changes, quality, safety, equipment delivery, BIM / IoT, and meeting records are usually stored in separate systems. Therefore, although the data are available, they may not provide timely decision support for project management. Traditional statistical models are not good at interpreting narrative evidence. When the project conditions or supply conditions change, the prediction accuracy of the static model may also decrease. In order to solve these problems, this paper proposes Power-LLM ValueLoop, a closed-loop framework for data value mining supported by large models. The framework identifies entities, events, causal signals, urgency, conflict evidence, and confidence levels through a pattern-based large model layer, and associates relevant results with structured data such as schedule, cost, resources, quality, and supply chain. Subsequently, the semantic-time model is used to estimate and calibrate the probability of major delays or failure events in the next 4 weeks. Confirmed outcomes, human corrections, time-decayed learning, drift monitoring, and threshold adjustment form the feedback loop. The system can deliver traceable risk early warnings, identify key risk drivers, and provide action recommendations for review by project personnel. This paper uses a synthetic project portfolio for evaluation, including 720 power engineering projects, 10 start-up batches, and 15,653 project-week observation records. Controlled concept drift was introduced from the 6th batch. In the four batches after drift, the macro average ROC-AUC of the feedback model was 0.773, the macro average PR-AUC was 0.366, and the Brier score was 0.105. At the recall-oriented operating point, the model identified 78.1 % of the risks in the next four weeks, with an average warning advance of about 9.1 weeks. For 80.2 % of the event items, the system can issue an early warning at least 4 weeks before the event. Compared with the static AI fusion model, the feedback mechanism improves the recall rate by about 0.140, but the false alarm rate also increases from 0.275 to 0.391. This shows that the improvement of model performance still needs to rely on reasonable threshold management and manual review. A supplementary Monte Carlo decision analysis examines prevalence and error-cost sensitivity at fixed operating rates. At the reported test prevalence, feedback has lower expected error loss only when the missed-event cost exceeds approximately 5.35 times the false-alarm cost. This conditional result does not establish savings in deployed projects.