May 2026
Cybersecurity challenges in industry 5.0: Human centric security, converged vulnerabilities, and resilience by design in smart manufacturing
2026, vol.18 , no.2, pp. 105-116
Article [2026-02-10]
Industry 5.0 redefines industrial transformation by complementing automation and connectivity with human-centricity, sustainability, and resilience. The increasing integration of artificial intelligence, industrial Internet of Things, collaborative robotics, cyber-physical systems, and cloud-edge infrastructures enhances manufacturing adaptability, but simultaneously enlarges the cyber-attack surface of industrial environments. Unlike conventional industrial settings, cyber incidents in Industry 5.0 may compromise not only data and control functions, but also operator safety, process integrity, and production continuity. This paper analyses the cybersecurity challenges of Industry 5.0 through three complementary dimensions: human-centric security requirements, vulnerabilities arising from IT/OT/IoT/AI convergence, and resilience-by-design protection strategies. The study provides a structured perspective for developing secure, trustworthy, and resilient smart manufacturing ecosystems.
Plamen Stanchev, Denitsa Ivanova-Mutafchieva, Nikolay Hinov. Cybersecurity challenges in industry 5.0: Human centric security, converged vulnerabilities, and resilience by design in smart manufacturing. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 105-116.
Simulation analysis of the structural impact of AI amplifiers on global risk maps
2026, vol.18 , no.2, pp. 95-104
Article [2026-02-09]
The aim of this paper is to present a simulation-based analysis of the structural impact of conceptual AI amplifiers on Global Cyber Risk Maps for 2023-2025. NetLogo was preferred by the author for the practical part of the study because of its advantages and suitability for creation of conceptual models of a wide variety of complex systems. The results are presented for different simulation scenarios, including one in which AI amplifies cyber protection and another in which it amplifies cyberattacks. Consequently, this experimental research contributes to a better understanding of how AI amplifiers affect cyberattack mechanisms and effective strategies for strengthened prevention.
Yoana Ivanova. Simulation analysis of the structural impact of AI amplifiers on global risk maps. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 95-104.
Comparative analysis of Paillier and BFV cryptosystems for building an electronic voting system
2026, vol.18 , no.2, pp. 83-94
Article [2026-02-08]
This study presents a comparative performance evaluation of the Paillier and BFV homomorphic encryption schemes applied to a simplified voting model using one-hot encoded ballots. The experimental setup involved a Python-based implementation of both schemes: a custom implementation of Paillier and the TenSEAL library for BFV. Key performance metrics, including encryption time, homomorphic aggregation time, decryption time, and ciphertext size, were measured across varying voter counts (1,000 to 10,000). Results demonstrate that Paillier has significantly lower aggregation time, while BFV achieves substantially lower encryption and decryption latency, especially under non-batching conditions. However, BFV offers stronger theoretical security guarantees under standard lattice assumptions.
Jalal Alizadeh, Ali Huseynli. Comparative analysis of Paillier and BFV cryptosystems for building an electronic voting system. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 83-94.
Privacy-robustness sensitivity analysis for secure development of differentially privacy deep learning under distribution shift
2026, vol.18 , no.2, pp. 71-82
Article [2026-02-07]
The widespread use of deep learning on sensitive data has positioned Differential Privacy (DP) as a central mechanism for protecting training confidentiality. Existing evaluations emphasize the privacy-utility trade-off measured on clean data, overlooking robustness under distribution shift. This study introduces Privacy-Robustness Sensitivity (PRS), a metric that explicitly quantifies the degradation in robustness induced by differential privacy by measuring the relative loss in corruption robustness with respect to clean accuracy as privacy constraints tighten. A privacy–utility–robustness evaluation framework is validated on CIFAR-10, CIFAR-10-C, CelebA, and a systematically constructed CelebA-C biometric corruption benchmark, enabling robustness-aware evaluation of privacy-preserving deep learning models for secure and resilient deployment scenarios.
B. Lavanya, S. Janani. Privacy-robustness sensitivity analysis for secure development of differentially privacy deep learning under distribution shift. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 71-82.
A lightweight transfer learning framework for real-time image classification in resource-constrained systems
2026, vol.18 , no.2, pp. 59-70
Article [2026-02-06]
This study investigates an EfficientNetB0-based transfer-learning approach for low-latency image classification on computing platforms with limited resources. The approach is evaluated on a four-class maize leaf disease dataset containing Blight, Common Rust, Gray Leaf Spot, and Healthy samples. The experimental protocol includes stratified dataset partitioning, training-time image augmentation, class-weighted optimization, and a two-stage training strategy in which the pretrained backbone is first frozen and then partially fine-tuned. Model performance is assessed using accuracy, macro-F1, balanced accuracy, ablation experiments, robustness tests, and deployment-related indicators. On the held-out test set, the model achieved 96.34% accuracy and a macro-F1 score of 0.95. Robustness evaluation under brightness changes, contrast shifts, Gaussian noise, and blur showed only limited performance degradation. The final model required approximately 28 MB of storage and achieved an inference time of about 3 ms per image. These results suggest that a compact transfer-learning model can support accurate and computationally efficient image classification in environments where memory, processing capacity, and response time are constrainеed.
Alketa Hyso, Dezdemona Gjylapi. A lightweight transfer learning framework for real-time image classification in resource-constrained systems. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 59-70.
Effective integration solutions for hybrid cloud migration to SAAS: A human capital management case study
2026, vol.18 , no.2, pp. 47-58
Article [2026-02-05]
The transition to a SaaS model often faces significant challenges, especially when integrating cloud-based solutions with existing established local systems, workflows, and data sources. This paper presents the results of a study related to the creation of flexible, scalable models and strategies for integration in hybrid cloud migration to SaaS. Through a case study of a company migrating its Human Capital Management infrastructure, we analyse the integration requirements necessary for a seamless transition. By systematizing these requirements and presenting practical design solutions, the study provides a framework for operational management to overcome real-world implementation hurdles. The research outcomes would significantly assist organizations to achieve full hybrid integration, ensuring both service continuity and long-term scalability during the move to the cloud SaaS model.
Nevena Staevsky, Rositsa Doneva, Silvia Gaftandzhieva. Effective integration solutions for hybrid cloud migration to SAAS: A human capital management case study. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 47-58.
Dynamic dictionary with conditional perfectly balanced binary search
2026, vol.18 , no.2, pp. 35-46
Article [2026-02-04]
In this paper we derived a novel dynamic dictionary set of algorithms that support perfectly balanced binary searches asymptotically. The dynamic dictionary is part of our FSP_vgv open-source C# package that aims to implement a portable version of Octave/Matlab in order to enable faster iterative design. Our package does not attempt to outdate other software, but fills specialized needs listed in this work. By processing the version control commit history of the FSP_vgv package we validate empirically that with unknown research horizon the time spent developing grows exponentially in the volume of production code. We identify an exponential growth parameter related to how intuitive a programming language is which also controls the research time hence proving the need for highly intuitive and portable languages.
Vladislav Vasilev, Georgi Iliev. Dynamic dictionary with conditional perfectly balanced binary search. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 35-46.
Effective load balancing in cloud computing using a hybrid optimization algorithm
2026, vol.18 , no.2, pp. 23-34
Article [2026-02-03]
Effective load balancing is a critical challenge in cloud computing environments, where dynamic and unpredictable workloads necessitate efficient resource allocation to ensure optimal performance and energy efficiency. This paper proposes a hybrid optimization algorithm, combining Quantum-Gray Wolf Optimization (Q-GWO) and Particle Swarm Optimization (PSO) to enhance load balancing in cloud computing. The Q-learning technique enhances the decision-making process by selecting appropriate virtual machines (VMs) based on their predicted fitness values. The hybrid GWO-PSO algorithm is then used to update positions and velocities, ensuring optimal resource utilization. Empirical evaluations demonstrate the superiority of the Q-GWO-PSO algorithm over traditional methods like Q-Learning Modified PSO (QMPSO) and Modified PSO (MPSO). Specifically, Q-GWO-PSO achieved a makespan of 7074.11 milliseconds (ms), throughput of 2129.16 requests per second (req/ms), a Standard Deviation (SD) of 0.05479, energy utilization of 150.979-kilo joule (KJ), and migrated 306 tasks, showcasing its robust and scalable solution for load balancing.
Anushree Kaushik, Gulista Khan, Priyank Singhal. Effective load balancing in cloud computing using a hybrid optimization algorithm. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 23-34.
BA-HGA: A burst-aware hybrid genetic algorithm for adaptive resource allocation in multi-tenant clouds
2026, vol.18 , no.2, pp. 11-22
Article [2026-02-02]
Multi-tenant cloud environments face unpredictable demand surges that challenge traditional scheduling algorithms. While recent AI-driven schedulers show promise, their behaviour under extreme burst conditions remains underexplored. This paper introduces BA-HGA (Burst-Aware Hybrid Genetic Algorithm), which integrates real-time burst detection through a Burst Intensity Coefficient (BIC), dual adaptive deadline policies, and dynamic fitness weight adjustment to maintain service-level agreement (SLA) compliance during workload volatility. We evaluate BA-HGA through 9,000 simulation experiments across varying burst intensities (0.1–0.9), comparing performance against classical schedulers (FCFS, Round-Robin) and state-of-the-art genetic algorithms. Results demonstrate that BA-HGA achieves 15–23% higher SLA success rates under high-burst conditions (probability ≥ 0.5) while maintaining comparable make span and load balance to conventional methods under stable workloads. The study contributes empirical evidence for burst-adaptive scheduling design and provides a practical framework for multi-tenant cloud resource management.
Alda Myzeqari, Jaumin Ajdari, Xhemal Zenuni. BA-HGA: A burst-aware hybrid genetic algorithm for adaptive resource allocation in multi-tenant clouds. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 11-22.
The task of cleaning up retrospective data on resource usage in the workload forecasting system of a distributed information system component
2026, vol.18 , no.2, pp. 3-10
Article [2026-02-01]
The article considers an approach to solving the problem of noise reduction of retrospective data on the level of utilization of computing resources of physical machines (using the example of processor cores) used as a training sample for a workload forecasting system of a virtualized data centre. The problems of noise reduction of retrospective data, such as the problem of "Noisy Neighbours" and Software Aging, are considered. A generalization of the subject area of data noise reduction for preprocessing time series of retrospective data based on signal processing methods, in particular, variants of mode decomposition methods, has been performed. Approaches to reducing the noise level of data based on the method of decomposition into variational modes (DVM) and variants of the method of decomposition into empirical modes (DEM) are considered. A hybrid scheme of the module for preprocessing a time series of retrospective data on the level of processor core utilization is proposed based on a comparative evaluation of the results of its variational mode decomposition and complementary multiple decomposition into empirical modes with adaptive noise in order to improve the quality of neural network forecasting of the workload of a virtualized data centre.
O. Ja. Kravets, Yu. V. Moiseeva, E. I. Mutina, O. Yu. Zaslavskaya, A. V. Gorshkov, Yu. V. Redkin, I. V. Atlasov. The task of cleaning up retrospective data on resource usage in the workload forecasting system of a distributed information system component. International Journal on Information Technologies and Security, vol.18 , no.2, 2026, pp. 3-10.