-
Research Article
AI-Driven Multi-Modal Vision Framework for Automated Detection and Quantification of Tube Blockages
Rahul Agnihotri*
,
Pallavi Wadhwa
Issue:
Volume 1, Issue 1, September 2026
Pages:
1-14
Received:
27 January 2026
Accepted:
26 August 2026
Published:
18 September 2026
Abstract: Heat exchanger tubes in power plants and refineries degrade through fouling, scaling, corrosion and wall thinning, and their internal condition governs both thermal efficiency and plant safety. Inspection of these assets is still performed largely by manual or semi-automated review of remote visual inspection footage, which is time-consuming, subjective, and prone to inconsistent defect interpretation between operators. The difficulty is compounded by the imaging environment itself: tube bores are narrow, illumination is supplied coaxially with the camera and falls off with depth, metallic surfaces produce strong specular reflections, and probe motion introduces blur. This work proposes an AI-driven multi-modal vision framework for automated detection, localization and quantification of tube blockages and surface degradation under these conditions. The framework integrates image preprocessing, deep feature learning, a hybrid convolutional neural network and vision transformer (CNN-ViT) detector, and a reinforcement learning agent that adapts probe traversal speed to the observed defect risk. A quantitative Tube Health Index (THI) is introduced to score severity from corrosion coverage, blockage ratio, defect depth and structural scaling, so that each tube is assigned an objective risk category rather than a binary defect flag. The framework was validated on an industrial dataset comprising 353 heat exchanger tubes, approximately 180 hours of footage and 12,400 labelled defect instances acquired under field conditions. The proposed hybrid model achieved 94.6% accuracy, 92.3% precision and 91.1% recall, outperforming conventional thresholding at 72.4% accuracy and a standalone CNN at 88.1% accuracy. Inference ran at 48 ms per frame on embedded hardware, and inspection time per tube fell by approximately 60% relative to manual practice. The results indicate that combining hybrid feature learning with adaptive scanning and quantitative health scoring offers a deployable pathway toward autonomous industrial non-destructive testing.
Abstract: Heat exchanger tubes in power plants and refineries degrade through fouling, scaling, corrosion and wall thinning, and their internal condition governs both thermal efficiency and plant safety. Inspection of these assets is still performed largely by manual or semi-automated review of remote visual inspection footage, which is time-consuming, subje...
Show More
-
Review Article
AI-Driven Insights for Detecting Corruption Risks in Ethiopian Public Sectors: A Systematic Literature Review
Tigist Mintesnot Tewabe*
Issue:
Volume 1, Issue 1, September 2026
Pages:
15-19
Received:
26 February 2026
Accepted:
9 March 2026
Published:
22 September 2026
Abstract: This systematic literature review examines AI-driven methods and evidence for detecting corruption risks in the public sectors of Ethiopia and neighboring East African countries. Corruption remains a significant challenge to effective public-sector governance, economic development, and the efficient delivery of public services in Ethiopia and East African countries. The increasing availability of digital government data and advances in artificial intelligence (AI) provide new opportunities to identify corruption challenges, detect anomalies, and strengthen transparency and accountability. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-guided search, screening, and selection procedures, relevant studies were identified from Scopus, Web of Science, PubMed, Google Scholar, and regional grey literature repositories. The review synthesizes evidence from 28 studies published between 2010 and 2025 that investigate the application of AI, machine learning (ML), network analysis, anomaly detection, and related computational approaches to corruption-risk identification. The review targeted on four major public-sector domains: public procurement, financial auditing, payroll management, and asset management. Key themes examined include the types and quality of data used, pre-processing and feature-engineering practices, dominant AI and ML algorithms, evaluation metrics, model interoperability, and the practical feasibility of implementation in low-resource environments. The results indicate growing interest in AI-enabled corruption detection, particularly through anomaly detection, classification, predictive modeling, and network-based approaches. However, the evidence base remains limited by fragmented datasets, insufficient data quality, limited access to government records, insufficient technical capacity, and the absence of standardized evaluation frameworks. Ethical concerns, including privacy, algorithmic bias, transparency, accountability, and the potential misuse of automated decision-support systems, also require careful consideration. In general, AI can complement, rather than replace, institutional anti-corruption mechanisms. For Ethiopia, priority should be given to strengthening public-sector data infrastructure, improving data governance and interoperability, developing interpret able and context-sensitive AI models, and establishing controlled pilot projects for procurement and financial-integrity monitoring. Sustained collaboration among policymakers, anti-corruption agencies, universities, technology experts, and international partners, together with capacity building and ethical safeguards, is essential for responsible and effective adoption of AI-driven corruption-risk detection systems.
Abstract: This systematic literature review examines AI-driven methods and evidence for detecting corruption risks in the public sectors of Ethiopia and neighboring East African countries. Corruption remains a significant challenge to effective public-sector governance, economic development, and the efficient delivery of public services in Ethiopia and East ...
Show More
-
Research Article
Design and Implementation of an Emergency Victim Localization and Rescue Coordination System for Disaster Response
Paulo John Kapembe*,
Zacharia Mzurikwao,
Candida Mwisoba,
Joseph Sospeter Salawa
Issue:
Volume 1, Issue 1, September 2026
Pages:
20-31
Received:
11 August 2026
Accepted:
25 August 2026
Published:
29 September 2026
Abstract: Communication failure during disasters can significantly delay the identification, localization, and rescue of affected individuals. When conventional communication infrastructure becomes unavailable or unreliable, victims may experience difficulties in communicating their emergency conditions and location to rescue personnel. This study presents the design and implementation of an Emergency Victim Localization and Rescue Coordination System for Disaster Response. The system was developed to provide a simple and accessible mechanism through which victims can submit emergency information and location data using a smartphone-based web interface, while rescue personnel can monitor submitted incidents through a centralized web dashboard. An ESP32 microcontroller is incorporated as the embedded control unit of the system and provides local system monitoring and status indication. The victim interface collects information including the victim's name, contact information, emergency category, description, and geographical location. Location information is obtained through browser-based geolocation services when permission is granted, while a map-based location-sharing option provides an alternative when automatic location acquisition is unsuccessful. Submitted emergency records are stored in a cloud database and synchronized with the rescue dashboard, where responders can view victim information, geographical coordinates, priority levels, and incident status. The prototype was implemented and evaluated through hardware testing, web-interface testing, database communication testing, and end-to-end emergency-reporting tests. The implemented system demonstrated the feasibility of integrating victim reporting, location acquisition, cloud-based data management, and rescue coordination into a single low-cost platform. The developed solution provides a practical approach for improving situational awareness and supporting faster coordination of emergency response activities, particularly in resource-constrained environments.
Abstract: Communication failure during disasters can significantly delay the identification, localization, and rescue of affected individuals. When conventional communication infrastructure becomes unavailable or unreliable, victims may experience difficulties in communicating their emergency conditions and location to rescue personnel. This study presents t...
Show More
-
Research Article
Forensic Analysis of Residual Bluetooth Artefacts in Android-Based Smartphones Before and After Device Unpairing
Issue:
Volume 1, Issue 1, September 2026
Pages:
32-39
Received:
5 June 2026
Accepted:
18 June 2026
Published:
29 September 2026
DOI:
10.11648/j.sdcomput.20260101.14
Downloads:
Views:
Abstract: Bluetooth technology is the most popular method of wireless communications in mobile phones, wearables, and Internet of Things devices. Bluetooth makes device connectivity and information transfer easy, and thus Bluetooth-related information can be useful digital forensic evidence. Research looking into the persistence of Bluetooth artefacts in Android smartphones after devices are unpaired is lacking. This study was designed to look into possible residual Bluetooth artefacts persistence and assess their forensic importance. A controlled experiment using Android devices was designed for the inclusion of Bluetooth artefacts through the pairing, connecting, and transferring of files. This was followed by manual device unpairing by the experimenter. Afterwards, Android Debug Bridge (ADB) was used to perform a logical acquisition of the paired Android device and extract paired Bluetooth information along with Bluetooth device identifiers, Bluetooth device MAC addresses, connection information, file information, and system information. Persistence of Bluetooth artefacts in the system before and after the unpairing of Bluetooth devices was assessed through a comparative analysis. The results of the study showed that unpairing Bluetooth devices does remove the bonding information, but residual artefacts still exist in the system. Evidence of user activities that includes transferred files, file information, and device identifiers was found to remain after the unpairing of the Bluetooth device. The transferred files along with the file information retained user file timestamps and showed a complete absence of changes, and thus provided evidence that the unpairing of the Bluetooth device did not remove all information connected to Bluetooth. Forensic evidence was found to be still available even after the active pairing(s) of the Bluetooth devices had been removed. These results are significant for forensic studies because they show how Bluetooth artefacts help to show how devices were used, what other devices were paired to, and what communication may have taken place. One of the contributions of this study to mobile and digital forensics is the understanding that the unpairing of Bluetooth devices is, in fact, not the full removal of evidence, but only the partial removal of evidence. This shows that forensic investigators should focus on the unpairing of Bluetooth devices to evidence.
Abstract: Bluetooth technology is the most popular method of wireless communications in mobile phones, wearables, and Internet of Things devices. Bluetooth makes device connectivity and information transfer easy, and thus Bluetooth-related information can be useful digital forensic evidence. Research looking into the persistence of Bluetooth artefacts in And...
Show More
-
Research Article
Efficient AI Architectures and Advanced Network Technologies for Future Intelligent Systems: The ICANOS Framework
Zubair Hussain*
Issue:
Volume 1, Issue 1, September 2026
Pages:
40-50
Received:
14 July 2026
Accepted:
24 July 2026
Published:
30 September 2026
Abstract: Transformer architectures have redefined the state of the art across natural language processing, computer vision, and signal processing, yet their quadratic computational complexity with respect to sequence length creates serious scalability and sustainability bottlenecks. At the same time, the convergence of sixth-generation (6G) wireless communication, the Internet of Things (IoT), and AI-native network intelligence introduces new complexity in resource management, proactive security, and energy governance. This paper argues that optimizing AI models or network infrastructure in isolation is insufficient: a tightly co-designed framework is needed that treats both concerns as inseparable. We survey efficient Transformer variants and hardware-aware approaches, and we characterize the networking landscape through 6G ubiquitous connectivity, AI-driven network automation, edge computing for IoT, and flexible network architectures. Drawing on these two bodies of literature, we propose the Intelligent Converged AI-Network Orchestration and Security (ICANOS) Framework, a three-layer AI-driven architecture comprising a Holistic Resource Orchestrator (HRO), a Proactive Security Manager (PSM), and an Energy Efficiency Optimizer (EEO). Unlike conventional software-defined networking controllers and network-functions-virtualization orchestrators, which treat AI models as opaque, schedulable workloads, ICANOS treats AI model behavior itself as a first-class signal for resource allocation and threat detection. Simulated benchmarks demonstrate energy savings of 57.9%, latency reduction of 63%, and a threat-detection improvement of 32.7% relative to non-integrated baselines. All experimental components use open-source, locally deployable models and publicly available datasets, and fully instrumented code with computational cost logging is provided for reproducibility.
Abstract: Transformer architectures have redefined the state of the art across natural language processing, computer vision, and signal processing, yet their quadratic computational complexity with respect to sequence length creates serious scalability and sustainability bottlenecks. At the same time, the convergence of sixth-generation (6G) wireless communi...
Show More
-
Review Article
The Security Duality of Large Language Model Agents:
A Systematic Review of Self-Security Protection and Cybersecurity Empowerment
Dasen Fu*,
Dalin Xiang
Issue:
Volume 1, Issue 1, September 2026
Pages:
51-56
Received:
11 September 2026
Accepted:
20 September 2026
Published:
30 September 2026
Abstract: Large Language Model (LLM) agents integrate memory, tool invocation, environment interaction, and multi-agent collaboration, evolving from passive text-generation tools into autonomous systems that execute complex tasks. This capability leap creates a pronounced security duality. On the one hand, LLM agents face emerging threats such as jailbreak attacks, prompt injection, memory poisoning, and multi-agent collusion, requiring systematic self-security protection. On the other hand, they empower cybersecurity in vulnerability detection, penetration testing, threat intelligence, and malicious code analysis, while their autonomous attack capabilities, e.g., zero-day exploitation, raise concerns about malicious abuse. Existing reviews often treat these two lines separately. This review systematically examines key studies from 2023 to 2026, covering attacks on agents and agent-enabled cyber operations. We propose a unified analytical framework with two dimensions: Dimension 1, self-security protection, and Dimension 2, cybersecurity empowerment. We organize self-security threats into model-layer safety alignment failure and fine-tuning-induced safety weakening; prompt-layer jailbreak and indirect prompt injection; memory/knowledge-layer retrieval-augmented generation (RAG) poisoning and long-term memory contamination; and multi-agent-layer prompt infection and secret collusion. For cybersecurity empowerment, we review vulnerability detection and remediation, automated penetration testing, cyber threat detection, autonomous vulnerability exploitation, and information ecosystem security. We further identify four coupling mechanisms: security weaknesses as attack entry points, tool integration expanding the attack surface, bidirectional capability enhancement, and double-edged multi-agent architectures. Finally, we distill five challenges—security-capability parity, native security design, multi-agent governance, ecosystem-oriented evaluation, and responsible capability release—and outline future directions for responsible development and deployment of LLM agents.
Abstract: Large Language Model (LLM) agents integrate memory, tool invocation, environment interaction, and multi-agent collaboration, evolving from passive text-generation tools into autonomous systems that execute complex tasks. This capability leap creates a pronounced security duality. On the one hand, LLM agents face emerging threats such as jailbreak a...
Show More
-
Research Article
Online Electricity Bill Management System: A Web Application for an Energy Supplier Company in Somalia Using Machine Learning
Mohamed Maawiye Hilowle*
Issue:
Volume 1, Issue 1, September 2026
Pages:
57-68
Received:
12 July 2026
Accepted:
3 August 2026
Published:
9 October 2026
Abstract: This paper presents the design and development of an Online Electricity Bill Management System (OEBMS) for an energy supplier company in Somalia, with the objective of improving the efficiency, accuracy, and reliability of electricity billing operations. The system is developed using the Waterfall model within the Software Development Life Cycle (SDLC), providing a structured framework for analysis, design, implementation, and testing. The proposed OEBMS replaces traditional manual and paper-based billing systems with a web-based application that enables customers to access billing services anytime and from anywhere. It offers key functionalities such as customer account management, electricity consumption monitoring, automated invoice generation, and secure online payment processing. A major contribution of this paper is the integration of machine learning techniques, including regression models and random forest algorithms, to analyze historical electricity consumption data and predict future usage patterns. This predictive capability supports energy providers in demand forecasting and resource planning while also enabling customers to better understand and manage their electricity usage. In addition, the system generates analytical reports that identify consumption trends, potential demand increases, and customer segments with varying usage behaviors. The system utilizes a MySQL database for efficient data storage and management and an Apache web server to support the application interface. By automating billing processes and reducing human intervention, the OEBMS minimizes errors, improves data accuracy, and reduces operational costs. Furthermore, it promotes environmental sustainability by eliminating paper-based processes. Overall, this paper demonstrates how the integration of web technologies and machine learning can modernize electricity billing systems and enhance service delivery in the energy sector.
Abstract: This paper presents the design and development of an Online Electricity Bill Management System (OEBMS) for an energy supplier company in Somalia, with the objective of improving the efficiency, accuracy, and reliability of electricity billing operations. The system is developed using the Waterfall model within the Software Development Life Cycle (S...
Show More
