The rapid advancement of quantum computing threatens to undermine classical cryptographic primitives, creating an urgent need for post-quantum secure alternatives. Lattice-based Attribute-Based Encryption (ABE) has emerged as a promising direction, combining fine-grained access control with quantum resistance derived from lattice assumptions such as Learning With Errors (LWE) and Ring-LWE. This paper presents a systematic literature review of lattice-based ABE schemes published between 2022 and 2025, following the PRISMA methodology to ensure a rigorous and transparent study selection process. From an initial corpus of more than 7,800 studies retrieved across IEEE Xplore, SpringerLink, Elsevier ScienceDirect, MDPI, and Google Scholar, 1,236 studies were screened, resulting in 90 works selected for in-depth qualitative and quantitative synthesis. The review analyzes the evolution of lattice-based ABE in terms of mathematical foundations, security models, performance characteristics, and application domains. Particular emphasis is placed on emerging use cases in cloud computing, the Internet of Things (IoT), healthcare, financial systems, and government compliance, each evaluated against requirements such as revocation support, lightweight design, searchability, and multi-authority governance. Quantitative synthesis reveals uneven feature coverage across the literature: searchability is supported in 45.6% of schemes and lightweight constructions in 36.7%, whereas policy hiding (16.7%) and revocation (22.2%) remain comparatively underexplored. Performance trends indicate ciphertext sizes averaging 5.8 kB per attribute, with only 12% of schemes achieving compact representations below 2 kB. Moreover, fewer than 36% of surveyed constructions provide IND-CCA security, highlighting persistent challenges in achieving robustness against adaptive adversaries. By integrating structured research questions with quantitative trend analysis, this study offers both descriptive and evidence-backed insights into the current state of lattice-based ABE. The findings identify critical gaps in efficiency, security guarantees, and standardization, and inform a forward-looking research roadmap emphasizing lightweight revocation mechanisms, efficient policy hiding, stronger IND-CCA security models, and the need for standardized benchmarking frameworks. This survey thus serves as a comprehensive reference for researchers and practitioners working on post-quantum secure access control systems, particularly in privacy-sensitive domains such as healthcare.

Large language models (LLMs) have achieved remarkable success on various natural language tasks, but their immense size often makes deployment on resource-constrained devices impractical. This paper presents Lightweight Nano LLM, a compact 125-million-parameter Transformer-based language model tailored for domain-specific question answering. Built on a GPT-2 small architecture, the proposed model is trained via knowledge distillation on a custom B.Tech CSE curriculum corpus ( $$\sim $$ 3 million tokens) to inject deep-domain knowledge. The study emphasizes techniques from prior literature, such as model compression and curated training data that enable small models to punch above their weight in coherence and accuracy. In evaluations, Lightweight Nano LLM demonstrates near-perfect accuracy on in-domain queries, outperforming other models of similar scale (e.g. GPT-2 small, DistilGPT-2, BERT-base) in this specialized task. However, the proposed model has outperformed all the above compared models with $$72.8\%$$ accuracy. The model’s compact size and focused training also allow it to run efficiently on consumer hardware (e.g. Apple M1), highlighting the promise of small LLMs for personalized and offline applications. The proposed work presents a detailed literature review, comparative analysis with related models, and an architectural diagram. The results show that with appropriate training data and design, smaller truly can be smarter in specialized settings for STEM disciplines.

Precision parameter extraction for Multi-Junction Solar Cells (MJSCs) is paramount for reliable simulation and performance optimisation, but is critically hindered by the models’ severe non-linearity and multi-modality. This research suggests implementing the Bonobo Optimiser Algorithm (BOA) to overcome this challenge. The core novelty lies in BOA's first application of high-precision parameter estimation for MJSCs, based on the Single Diode Model (SDM), Double Diode Model (DDM), and Triple Diode Model (TDM). Additionally, statistical validation using ANOVA and the Friedman test confirms the effectiveness of the BOA for parameter estimation of the Thin triple-junction CTJ30 and multi-junction solar cell, as well as the MJSC, under various operating temperature conditions (41.5 °C, 51.3 °C, and 61.6 °C). Minimisation of RMSE is an objective function to optimise parameters. Statistical analysis of 30-run simulations confirms BOA’s superior consistency and minimal variance. Critical findings show that for the CTJ30 cell, BOA attained a consistent best RMSE of 2.513060E-03, 2.510171E-03 and 2.510387E-03 across SDM, DDM, and TDM, respectively. For MJSC datasets at 41.5 °C, BOA achieved the best RMSE values of 4.361990E-05 for the SDM, 2.768602E-05 for the DDM, and 2.827875E-05 for the TDM. At 51.3 °C, the best RMSE values were 5.434700E-05 for the SDM, 2.641600E-05 for the DDM, and 2.347326E-05 for the TDM. At 61.6 °C, the best RMSE values of 1.657126E-05, 2.693035E-05 and 1.556626E-05 for the SDM, DDM, and TDM, respectively, demonstrate exceptional precision. Empirical investigations and comparisons with the most recent state-of-the-art algorithms and conventional algorithms are used for validation and justification. Statistical ANOVA and the Friedman test validation and comparison investigations demonstrate the suggested BOA precision and consistency, which may replicate experimental datasets with minimal statistical lapses and reduced variance. A comprehensive statistical analysis of the CTJ30 and MJSCs PV datasets demonstrates that the proposed BOA is reliable in assisting quality control in high-efficiency photovoltaic systems.

Feature selection is an important preprocessing step in classification. By reducing dimensionality and removing irrelevant features, it can help improve classification performance. Swarm intelligence–based metaheuristic algorithms are widely used for feature selection. They are effective for exploring large and nonlinear search spaces. However, they often function as black-box approaches, providing limited transparency regarding the selection of particular features. The growing emphasis on eXplainable Artificial Intelligence has increased the need for interpretable decisions in sensitive domains such as healthcare and finance. This study focuses on improving the Firefly Algorithm for feature selection. The goal is to make the selection process easier to explain. It integrates SHapley Additive exPlanations values into the fitness function. The basic Firefly Algorithm is prone to premature convergence and local optima entrapment as it follows a single-leader attraction model. The proposed variant overcomes these limitations while maintaining a balance between exploration and exploitation. Experiments conducted on ten benchmark datasets show that the proposed approach achieves higher classification accuracy, more stable convergence, and feature subset sizes comparable to those of existing methods. The selected features also show strong overlap with SHAP-based importance rankings, enhancing the transparency of the feature selection process.

Precise segmentation of lung nodules is critical for early diagnosis and treatment of lung cancer. However, traditional methods are prone to boundary missegmentation and missed detection owing to the irregular shapes of nodules, blurred edges of small lesions, and vascular adhesion interference in CT images. This study proposes an improved U-Net model, SWAG-Net, which enhances the segmentation accuracy of edge-complex nodules and small-sized nodules through multi-module collaboration. The model designs a bidirectional path aggregation network that fuses cross-level feature pyramids to integrate the deep semantic information from the encoder and shallow details from the decoder, thereby enhancing the feature representation of small lesions. It embeds an adaptive multi-subband wavelet convolution module, utilizing wavelet transforms to explicitly separate multi-scale features and adaptively enhance high-frequency detail subbands, thereby addressing edge blurring and adhesion interference. Additionally, it introduces a dual-domain dynamic gating attention mechanism that combines channel-domain semantic filtering and spatial-domain edge enhancement to dynamically suppress background noise and highlight nodule regions. In the preprocessing stage, coordinate system decoupling and adaptive normalization were employed to reduce data bias. Experiments on the LUNA16 dataset demonstrated that the model achieved a DSC of 93.37% and an IOU of 88.79%, representing improvements of 11.43% and 19.39%, respectively, over the baseline U-Net. These results significantly improved the segmentation performance for irregular and small nodules. The experimental results demonstrate that this model effectively reduces mis-segmentations and missed detections, providing a reliable auxiliary basis for the early diagnosis of lung cancer.

A prediction method based on machine learning neural network model is proposed to address the problem of difficult prediction of concrete strength due to the interaction of multiple factors Historical data containing various influencing factors and corresponding strength values are collected and preprocessed, key features are extracted as input variables. A BP neural network is used to construct a model to handle complex nonlinear problems, the network structure is adjusted to adapt to the task, and weights and thresholds are optimized through forward and backward propagation algorithms. Use the trained model for a new task and input the predicted values of the new data. The experiment shows that the model performs well in predicting the compressive strength of concrete at 7 and 28 days, with an R² of 0.98 at 28 days and 0.95 at 7 days, demonstrating high accuracy and practicality.

The increasing prevalence of advanced technology in sports has empowered athletes, coaches, and team management to elevate their performance and fine-tune team tactics. Notably, player collaborations have emerged as a critical factor in reshaping the dynamics of matches, yet a lack of tools for accurate partnership analysis often leads to instinct-driven decisions rather than data-based insights. The aim of this study is to investigate the potential of harnessing machine learning algorithms to evaluate the significance of player partnerships in football and their potential impact on match out-comes. This study innovatively forecasts player partnership formations in live matches, demonstrating adept management of extensive, unstructured datasets to create a robust solution. It offers valuable insights for coaches to predict and build effective player partnerships, resulting in sustainable advantages for the team. The research utilizes machine learning techniques, including Logistic Regression (LR), Decision Tree (DT), K-nearest Neighbor (KNN), Support Vector Machine (SVM), Multinomial Naïve Bayes (MNB), and Random Forest (RF) classifiers. These models were applied and evaluated to construct models and assess player partnerships. Necessary data was sourced from the Kaggle dataset, which encompasses information on 167,798 player assists from 9,074 European league football matches spanning six seasons across five leagues. Machine learning models underwent training and evaluation through an 80 − 20 data split and further verified through cross-validation to ensure robustness across league datasets. Balanced sampling and k-fold cross-validation were used to mitigate data imbalance and validate model generalization. Notably, the SVM, DT and RF models demonstrated impressive accuracy of 99%, with the highest F1-score of 0.99, encompassing all football leagues. The SVM and RF models also achieved the highest precision, at 0.99, albeit with a relatively longer computation time of up to 60 s. Model performance varied marginally across leagues (± 0.3% standard deviation in accuracy and F1-score), indicating stable predictive behaviour across gameplay variations. Unlike prior studies that analyzed team synchronization or motion dynamics, this work uniquely models player partnerships based on assist-goal frequency patterns and forecasts their continuity across matches using machine learning classifiers. This study provides valuable insights for devising strategies that can confer a competitive edge in matches. The findings offer a practical means for team managers and coaches to make well-informed decisions regarding player partnerships.

Abstract Induction motors play a critical role in industrial operations and electric vehicles, yet their reliability is often compromised by stator winding inter-turn short-circuit (ITSC) faults. To mitigate costly downtimes, this study investigates machine learning-based fault detection methods, comparing Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM) networks, and Physics-Informed Neural Networks (PINNs). A benchmark dataset comprising stator current signals, leakage flux, and varying load conditions was employed to train and evaluate the models. While classification across seven fault categories remained challenging, the proposed framework achieved outstanding results in binary classification (healthy vs. faulty). In particular, the LSTM model attained a near-perfect accuracy of 100%, and the PINN model achieved 99.32%, both surpassing the baseline ANN performance of 99.01%. These findings highlight the effectiveness of temporal modeling with LSTM and the added value of incorporating motor physics into PINNs, especially in scenarios with limited data. The results confirm that advanced machine learning models can significantly improve early fault detection, enabling more reliable predictive maintenance and reduced operational costs in industrial systems.

Swarm robotics is an emerging field capable of accomplishing complex tasks through collective behaviour. However, it continues to face persistent challenges in secure communication, decentralized decision-making, and scalability. To operate effectively in resource-constrained environments, swarm networks require a decentralized mechanism that is secure, fast, and efficient. Although many studies have explored the use of blockchain technology for swarm robotics, existing blockchain consensus algorithms such as Proof of Work (PoW), Proof of Stake (PoS), and their variants remain unsuitable due to high computational complexity and risk of stake centralization. To address these challenges, we introduce the blockchain-based Rotational Leadership Role (RLR) consensus algorithm, a voting-based consensus re-engineered from the Raft approach, together with Decentralized Task Authorization and Validation (DeTAV), a token-based mechanism for context-aware task validation. This design ensures efficiency, security, and scalability in swarm robotics and drone systems. RLR is lightweight and well suited to operate within the limited computing resources of small robots or aerial drones. To validate its performance, a custom-built robotic simulator was developed as part of this research. Experiments conducted with up to 70 concurrent robots demonstrated that RLR consumed under 90 MB Random Access Memory (RAM) and 12% Central Processing Unit (CPU), whereas PoW required 460 MB RAM and 27% CPU with a minimum difficulty level of 21, reflecting an 80% reduction in memory usage and a 55% reduction in CPU consumption. Scalability tests with 4 to 70 robots further revealed RLR’s scalability with an average of 78% higher throughput, 47% lower election latency, and 34% lower consensus latency. Additionally, under the simulated attack scenarios and assuming uncompromised cryptographic keys, DeTAV’s context-based validation consistently achieved 100% success in detecting and isolating Byzantine nodes, while reducing Quality of Detection (QoD) time by 67%. Collectively, these results confirm that RLR with DeTAV effectively meets the efficiency, security, and scalability requirements of swarm robotic and drone networks.