Clustering in wireless sensor networks (WSNs) offers numerous desirable properties, including load balancing, energy conservation, and distributed key management. Secure Clustering requires it to detect compromised nodes and remove them from clusters during setup. Suppose some nodes are attacked and pass the filtering. In that case, they can modify some nodes to adopt a different clustering perspective, as well as initiate new clusters to degrade the overall cluster quality. To address these issues, a new method, Secretary Bird with Self-Organizing Maps (SBWSOM), has been designed to detect and eliminate malicious nodes while efficiently providing data. First, the appropriate sensor nodes were constructed in Python. Second, the malicious node was located and destroyed, and the Cluster Head (CH) was picked based on parameters such as remaining energy, network level, and base station (BS) location. Furthermore, the data rates of chosen CHs have been confirmed and sent to empty nodes. Lastly, the values compared and studied were Latency, throughput, packet delivery ratio (PDR), energy consumption, and transmission loss. The evaluation of this proposal demonstrated improved data transfer, with a throughput of 0.91, an energy consumption of 0.46 mJ, and a packet delivery ratio of 96.3%. Also, the transmit loss was 4.20%, and Latency was 6.04 ms. Overall, this method performed well, with significant improvement over previous models.
Driver behavior, vehicle dynamics, and operating conditions strongly influence fuel consumption in open-pit mining operations. This study proposes a driver behavior–aware fuel optimization framework that integrates a digital twin architecture with reinforcement learning to improve fuel efficiency of heavy-duty haul trucks. The framework combines a data-driven vehicle dynamics surrogate, explicit modeling of driver behavior, and proximal policy optimization to enable safe and scalable policy learning within a realistic simulation environment. Historical telematics data were used to construct the digital twin and evaluate the learned policy under controlled operating conditions. Experimental results show that the reinforcement learning agent produces substantially smoother driving behavior, characterized by stable speed regulation and elimination of aggressive acceleration and braking events. Compared to historical operator driving, fuel consumption per kilometer, computed using rollout-level aggregation of cumulative fuel consumption and total traveled distance, was reduced from 4.45 L/km to 3.02 L/km, corresponding to a 32.05% improvement in fuel efficiency. The results demonstrate that explicitly modeling driver behavior within a digital twin-based reinforcement learning framework can yield significant fuel savings while maintaining realistic and interpretable driving strategies. The proposed approach provides a promising foundation for the development of decision-support and driver assistance systems aimed at improving energy efficiency in open-pit haulage operations.
This paper presents a fully integrated, real-time assistive system that combines voice-based object recognition with a generative conversational interface, specifically designed to enhance elderly care through edge AI deployment. The proposed framework enables intuitive human–robot interaction in domestic environments by fusing natural language understanding, optimized visual detection, and local generative response. Voice commands are processed through a speech-to-text pipeline using the Google Web Speech API, with keyword extraction triggering object detection via a quantized YOLOv8n model accelerated through TensorRT with FP16 inference on an NVIDIA Jetson Nano. In parallel, a locally deployed generative AI assistant, executed entirely on-device, provides empathetic dialogue to support social engagement and emotional well-being. The proposed system adopts a hybrid edge architecture in which object detection, robot control, and LLM-based dialogue generation are executed on-device, while speech-to-text transcription relies on a cloud-based service. This generative interface is implemented as an LLM-based Emotional Assistant. The system achieves 13 FPS with an inference latency of 70 ms for object detection, 94.3% speech recognition accuracy, and an F1-score of 0.69 at a 0.5 confidence threshold. All AI components are executed on-board, preserving privacy for on- device processing while maintaining real-time responsiveness. Experimental validation confirms the effectiveness of deploying multimodal AI, including generative models, on resource-constrained hardware. This work lays the foundation for autonomous, voice-guided care robots that not only assist in locating objects but also engage users socially, promoting greater autonomy and quality of life for older adults.
Blockchain technology has emerged as a transformative tool for secure data sharing across decentralised systems, particularly in finance, healthcare, and governance. However, despite its promise, the widespread adoption of blockchain platforms remains constrained by unresolved security threats and architecture-specific vulnerabilities. This paper presents a systematic literature review (SLR) that critically evaluates the security risks and countermeasures associated with blockchain-based data sharing models. The review focuses on three widely referenced platforms, Ethereum, Hyperledger Fabric, and MedRec, chosen for their relevance to public, permissioned, and healthcare-oriented blockchain deployments, respectively. The review analyzed 30 peer-reviewed publications from 2018 to 2025 sourced from IEEE Xplore, SpringerLink, ScienceDirect, and other digital libraries. Empirical insights from the reviewed literature indicate that Sybil attacks remain prevalent on public blockchains, although adaptive Proof-of-Stake protocols are reported to reduce their success rate considerably. Front-running and Miner Extractable Value–related behaviors are frequently observed in Ethereum-based decentralised finance ecosystems, often resulting in significant financial losses. Unauthorised access persists as a major concern, particularly for software wallets, which are commonly exposed to phishing and malware attacks. The findings underscore unique trade-offs across platforms: Ethereum supports transparency but is prone to transaction manipulation; Hyperledger ensures strong access control yet faces insider threat challenges; and MedRec enhances patient privacy but lacks robust mobile integration. This study provides a structured synthesis of existing threats, platform-level responses, and design trade-offs, offering guidance for stakeholders aiming to strengthen security in blockchain-based data-sharing infrastructures.
Network Intrusion Detection Systems (NIDS) play a vital role in modern cybersecurity by leveraging artificial intelligence (AI) in particular deep learning (DL) and machine learning (ML) to detect and mitigate malicious activities. However, these AI-driven systems are highly vulnerable to adversarial attacks, where small, imperceptible perturbations in input data can deceive models and significantly reduce detection accuracy. This raises critical concerns about the security and reliability of intrusion detection, especially in real-world scenarios where attackers exploit adversarial transferability to bypass defenses. This research investigates the threat posed by black-box adversarial attacks via surrogate models, focusing on the ability of adversarial examples to transfer across different architectures. This study simulates real-world adversarial threats, demonstrating how attacks crafted on one model can effectively deceive another, compromising NIDS security. A comparative study is conducted on two widely used AI models: an Artificial Neural Network (ANN) and a Convolutional Neural Network (CNN), both trained on the CICIDS 2019 dataset. The study evaluates the robustness of these models against two gradient-based adversarial attack methods, Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD), to determine their susceptibility under black box adversarial conditions. Experimental results indicate that CNN-based NIDS are more vulnerable to adversarial attacks than ANN-based models, with adversarial examples successfully transferring across architectures. These findings highlight the critical risks associated with adversarial transferability, underscoring the need for enhanced security measures to strengthen AI-driven intrusion detection systems against evolving cyber threats.
The Village Credit Institution (LPD) is a microfinance institution that plays a vital role in the rural economy in Bali. LPDs provide credit through a manual and subjective analysis process conducted by loan officers. This often hinders the objectivity and consistency of credit analysis. Modernization is a strategic step to address this issue. This study aims to analyze the digital transformation process in the LPD credit analysis system through the implementation of machine learning. Initial analysis indicates that debtors with long tenors, high delinquency rates, and high debt-to-income ratios have a higher risk of default. This pattern serves as the basis for learning a machine learning model using Logistic Regression, Decision Tree, Random Forest, and XGBoost algorithms to classify creditworthiness. The XGBoost algorithm demonstrated the best performance with an accuracy of 93% and an AUC of 0.96. Regarding credit approval, this model was able to identify high-risk potential debtors with significantly better accuracy than conventional methods. With the ability to process thousands of historical data points in a relatively short time, thereby accelerating decision-making, the application of machine learning significantly improves the efficiency and objectivity of credit analysis. This supports the realization of digital transformation in credit analysis at LPDs.
The growth of Internet of Things (IoT) networks has drastically improved attack surface, requiring intrusion detection systems (IDS) to ensure accuracy and privacy protection. To overcome these obstacles, we introduce a federated learning (FL) based IDSW that incorporates state-of-the-art preprocessing, smart feature optimization, and a new classification paradigm. During preprocessing, raw traffic data is subject to scrubbing at a vigorous level, normalization through scaling, and label encoding to maintain consistency and reduce noise in heterogeneous local datasets. For feature selection, the Hybrid Emperor Penguin–Quokka Swarm Optimization (HEPQSO) approach is utilized which balances exploitation and exploration to find the most discriminative features while addressing the dimensionality problem. These features are then utilized by a deep hybrid classifier where the Spike Gated Linear Unit (SGLU) facilitates non-linear representation learning, and a Vision Transformer-Temporal Convolutional Network (ViT–TCN) hybrid discovers both global spatial relationships and local temporal dynamics of intrusion patterns. Experimental analyses performed using benchmark intrusion detection datasets show that the system has a high performance compared to baseline models at all times, with an accuracy of 97.88%, precision of 96.16%, recall of 97.54%, F1-score of 97.39%, specificity of 97.62%, and MCC of 97.04%, thus proving its efficiency for safe IoT settings. This combination of state-of-the-art preprocessing, hybrid feature selection, and deep federated classification forms a robust IDS that can tackle the changing landscape of cyber intrusions.
In high-hazard workplaces like packaging facilities, effective fire safety is critical, but conventional practices fail to recognize actual hazards and are highly expensive to implement. This paper presents a hybrid reconstruction and artificial intelligence-driven framework that can potentially be applied to build interactive virtual reality environments. The objective of this study is to develop a scalable and cost-effective Virtual Reality based fire safety training system that balances realism and interactivity. To balance visual fidelity and interactivity, a hybrid reconstruction pipeline was developed. The complex background environment was reconstructed and rendered using 3D Gaussian Splatting, while for reconstructing key industrial objects as solid and interactive meshes, photogrammetry is used. An artificial intelligence-based system has been adopted for automatic object detection using You Only Look Once version 11 (YOLOv11) and material-based hazard classification using Bidirectional Encoder Representations from Transformers (BERT). In addition, interaction options are generated using a text generation model Fine-tuned Language Net Text-to-Text Transfer Transformer (FLAN-T5). The results indicate that the proposed framework produces high rendering capabilities with high precision, enabling efficient and scalable development of industrial safety training modules.
Depression is a serious psychiatric disorder that greatly impacts the quality of life and daily functioning of a person. Accurate diagnosis at the early stage is critical for success with intervention. Electroencephalography (EEG) offers a non-invasive technique to assess neurophysiological activity and is thus an important instrument for diagnosis of depression. Current EEG-based deep learning approaches are beset by high-dimensional data, poor feature selection, and poor classification performance owing to the nature of the EEG signal. To address these issues, we introduce EEGEffV2-SpikeNet a new framework for depression detection that combines statistical feature extraction with deep feature extraction through a Graph Convolutional Network (GCN) approach. The proposed model incorporates a new fusion of statistical feature extraction and GCN-based deep feature learning for the extraction of both spatial and temporal EEG features. The extracted features are then optimized by Modified Addax Optimization Algorithm (MAOA), which is a cutting-edge bio-inspired optimization algorithm for optimizing feature selection efficiency by discarding redundant information and improving classification accuracy. For depression classification, EfficientNetV2, Deep Belief Network (DBN) and a Spiking Neural Network (SNN) are utilized based on the computational efficiency of EfficientNetV2 and the biologically simulated processing of SNN for enhancing feature representation and decision-making. Experimental results on two standard EEG datasets validate the better performance of the model, achieving 98.74% accuracy on Dataset 1 and 97.88% accuracy on Dataset 2, outperforming baseline models like DBN, EfficientNet, and SNN. The results prove the framework's promise as a dependable tool for objective and early depression diagnosis, with clinical application and mental health monitoring implications.
Machine learning (ML) has made it much easier to find and estimate the risk of early stage of cardiovascular illnesses by making it possible to analyses massive, various clinical datasets quickly and easily. In these kinds of datasets, demographic information, lifestyle characteristics, medical history, and diagnostic measurements are all included. These are all things that may not be easy to see through standard clinical examination. This study examines heart disease prediction through a series of hybrid ML models that integrate neighborhood-based classifiers, swarm intelligence-driven optimization, and ensemble learning, motivated by existing obstacles. There are four hybrid models being proposed: MSMO-KE and MSMO-KM, which combine Modified Spider Monkey Optimization (MSMO) with K-Nearest Neighbour classifiers that use Euclidean and Minkowski distance measures, respectively. There are also two ensemble variants, MSMO-KECB and MSMO-KMCB, which add CatBoost as a final prediction layer. To make sure it is strong and can be used in other situations, the proposed framework is tested on three separate cardiovascular datasets using a cross-validation method. The experimental findings show that the performance is always better than the baseline and the best models that are already used. The MSMO-KMCB model performs the best overall out of all the approaches tested. It has a cross-validated accuracy of 98.2% on Dataset-3 while keeping a high sensitivity. The comparative research demonstrates that the proposed MSMO-based ensemble models surpass current methodologies in predictive accuracy and recall, underscoring their promise for dependable and efficient heart disease risk prediction in clinical decision-support systems.
Education is crucial for personal and economic growth, but financial challenges in developing countries hinder equitable academic success. NGOs administer scholarship programs to empower underprivileged individuals, a crucial step towards the attainment of Sustainable Development Goal 4, which aims to provide inclusive and equitable quality education for all. This study proposes a novel Scholarship Award Recommendation System that leverages predictive modelling and ensemble learning to identify deserving students for scholarship awards. The system utilizes a robust ensemble model that combines the strengths of Quadratic Discriminant Analysis (QDA), Random Forest (RF), and Extra Trees (ET) to predict students' academic performance. Additionally, we incorporate answers from the General Mental Health Questionnaire (GHQ-12). The GHQ-12 responses are pre-processed using a binary scoring approach (0-0-1-1) and integrated as predictive variables alongside academic and demographic features. We apply this framework to a case study of Nigerian university students in partnership with Springtime Development Foundation. The results indicate that incorporating GHQ-12 features significantly enhances prediction accuracy, with QDA, RF, and ET achieving accuracy scores of 0.90, 0.86, and 0.89, respectively. Statistical analysis using a t-test confirms the relevance of GHQ-12 features, with a p-value of 0.0013 establishing a significant correlation between student performance and mental health status. The study showed the effectiveness of the ensemble model to accurately predict students’ academic performance. It highlights the significance of incorporating variables from the (GHQ-12) into the predictive model, indicating mental health as a crucial factor for predicting academic performance which in turn enhances the performances of the Classification Models considered.
The rapid integration of conversational Artificial Intelligence (AI) into instant messaging platforms has transformed how younger generations interact with digital technology. This study investigates Generation Z's engagement with Meta AI on WhatsApp by employing a modified Technology Acceptance Model (TAM) that partitions End-User Computing Satisfaction (EUCS) dimensions to ensure measurement validity. Specifically, 'content' and 'accuracy' reflect Perceived Usefulness, while 'format' and 'timeliness' reflect Perceived Ease of Use. A quantitative survey involving 272 Generation Z respondents in Indonesia was analyzed using Covariance-Based Structural Equation Modeling (CB-SEM) with AMOS. The results reveal that Perceived Ease of Use significantly influences both Perceived Usefulness and Attitude Toward Using. Consequently, Perceived Usefulness acts as a partial, rather than full, mediator between ease of use and user attitude. Furthermore, Attitude Toward Using emerged as a powerful determinant of Actual System Use, with the proposed model explaining 89% of its variance (R2 = 0.89). These findings suggest a synergistic effect for "digital natives": while an intuitive format and fast response times directly foster positive attitudes, the epistemic quality and accuracy of the AI remain the dominant drivers of sustained engagement. This study contributes theoretically by validating a robust, multicollinearity-resistant modified TAM for conversational AI, providing practical insights for developers to maintain frictionless interfaces while prioritizing algorithmic accuracy to enhance user adoption.
Part-of-Speech (POS) tagging is an essential and important pre-processing activity for many Natural Language Processing (NLP) applications, this is particularly more evident for morphologically rich languages such as Marathi. This research investigates POS tagging for Marathi using the Maximum Entropy Markov Model (MEMM). MEMM combines the strengths of conditional probability modelling and sequence prediction, allowing the integration of rich contextual features. Features used include word forms, suffixes, prefixes, and neighboring tags, effectively tackling the challenges presented by inflectional variations and ambiguity in Marathi. Experimental results demonstrate that the MEMM-based POS tagger achieves an accuracy of 83.72%. This performance marks a notable advancement in Marathi POS tagging, given the linguistic diversity and the scarcity of annotated data. Error analysis enhances the issues like ambiguity in homonyms and out-of-vocabulary words, providing methods for further improvement through enriched datasets and sophisticated modelling techniques. This study enhances NLP applications such as machine translation, spell checking, and sentiment analysis for Indian languages and offers a solid foundation for future research in Marathi POS tagging.
Requirements change management is one of the core business analyst's activities, directly affecting change impact analysis, stakeholder communication, and the long-term system maintainability. While research on this topic examines in detail change processes, tracking methods, and change type classification, the problem of systematically documenting requirements changes remains underexplored. Existing research lacks a unified classification of change documentation approaches and context-sensitive recommendations for their selection, which limits their effectiveness in managing requirements. To address this gap, this study develops a context-oriented framework for selecting approaches to requirements change documentation. The framework integrates three components: a conceptual model based on the Baseline–Delta–Target State triad, a taxonomy of documentation approaches, and a context-driven selection mechanism grounded in empirical evidence. A systematic literature review was combined with an analysis of the survey of 324 practicing business analysts from Ukrainian and international companies. Statistically significant associations between selected project context attributes and documentation practices were identified using the Chi-square test of independence and Cramer's V, while additional dimensions were supported through evidence from the literature. The framework incorporates six documentation approaches: Full Target State, Delta-only, Target-driven Delta, Delta-driven Target, Parallel Use, and Hybrid Cycle. Four contextual dimensions emerge as key factors: project, environment, resources, and stakeholders. To support context-based selection of the change documentation approach, a matrix was developed that integrates the identified dependencies. The results position requirements change documentation as a context-sensitive knowledge management mechanism rather than a universal procedural standard.
Sinusitis is an inflammation of the paranasal sinus mucosa, which is an infection caused by a bacterium, fungus or virus. Therefore, for earliest and accurate prediction of sinusitis from Computed Tomography (CT) image, this research introduces a novel Artificial Intelligence (AI) based technique. The developed research is initiated with preprocessing using a Gabor filter to improve the quality of an image. After, segmentation using Gaussian Mixture Model (GMM) is exploited for effective isolation of sinus regions affected by inflammation. For acquiring the crucial features from the segmented regions, Gray-Level Co-occurrence Matrix (GLCM) based feature extraction is utilized which offers clinically meaningful features that improve transparency. Consequently, the hybrid Harmony Search Algorithm (HSA)-Grey Wolf Optimizer (GWO) feature selection is utilized to choose the most relevant features. This hybrid method outperforms traditional selection techniques by effectively identifying the most discriminative and non-redundant features, enhancing classification accuracy while reducing computational complexity. For accurate classification of sinusitis into various severity levels, the modified Artificial Neural Network (ANN) is employed. Unlike end-to-end deep learning models, this modular approach allows for fine-grained control at each stage, ensuring that critical medical insights are not lost in abstraction. This structured pipeline allows each phase to be optimized individually, improving transparency, reliability and ultimately, diagnostic performance. The performance of the research is analyzed via python software and it reveals that the developed classifier achieves an accuracy of 96.41%.
With the rapid proliferation of electronic devices, the volume and sophistication of malware have surged, posing critical cybersecurity threats. Traditional malware detection approaches face challenges such as limited generalization, unbalanced datasets, and high computational costs. To address these issues, this study introduces the LLM-Powered Transformer Framework for Multi-Class Malware Detection, an image-based approach integrating Large Language Models (LLMs) and transformer architectures with Convolutional Neural Networks (CNNs). The proposed framework enhances malware classification by leveraging data visualization, balanced sampling, and data augmentation techniques, achieving over 98.86% accuracy across four open-source datasets. Furthermore, this study makes two key contributions: first, it provides granular insights into malware classification performance using confusion matrix analysis, aiding cybersecurity professionals in refining detection strategies. Second, the balanced sampling approach eliminates the need for additional datasets, minimizes hardware overhead, and dynamically adjusts sampling weights for optimal learning. Additionally, data augmentation techniques mitigate overfitting, enhancing the model's adaptability to diverse malware variants. Comparative analysis with state-of-the-art methods demonstrates the proposed framework's efficiency in achieving high accuracy while maintaining computational feasibility. These advancements establish a robust foundation for real-world malware detection and cybersecurity applications.
As classical computation approaches its fundamental limits related to power dissipation, reversible logic, which theoretically achieves zero energy loss, is becoming a critical technology for future low-power and quantum computing. However, most research in this field remains theoretical, lacking practical, hardware-verified implementations. This paper bridges this gap by presenting the complete hardware implementation and rigorous fault-tolerance validation of a reversible encryptor based on extended Fredkin gates. First, we detail the full realization of the encryptor on an Altera Cyclone IV Field-Programmable Gate Array. This implementation is not just a simulation but a complete, interactive hardware prototype, featuring real-time data input via a standard keyboard and output to a video graphics array monitor. Second, since functional verification is insufficient for cryptographic hardware and exhaustive testing is computationally infeasible, we introduce a novel validation methodology. This core contribution utilizes a metaheuristic ant colony optimization algorithm, not for synthesis, but for the intelligent generation of an optimal and compact set of test vectors. This test set is designed to achieve maximum fault coverage for the industry-standard "stuck-at fault" model. The algorithm successfully generated a minimal test set achieving 100% coverage for the considered single stuck-at fault model. We then experimentally validated this methodology by manually injecting a stuck-at fault into the hardware description language, recompiling the faulty circuit into the device, and confirming that the metaheuristically generated test vector successfully detected the physical fault. Thus, this work demonstrates the full cycle from theory to a practically validated and reliable hardware implementation of a reversible system.
Recommender system commonly suffers from data sparsity and cold-start problems, where user-item interactions hinder reliable preference learning. While recent Graph Neural Network based models such as LightGCN and NGCF effectively capture higher-order collaborative signals, they primarily rely on interaction-derived embeddings and remain sensitive to sparse environments. This paper Attribute Enabled Graph Neural Framework (AE-GNF) proposes a Semantic-Aware Graph Refinement Framework that integrates attribute-driven representation learning with graph-based collaborative propagation to address these limitations. The proposed method first encodes heterogeneous user and item attributes using semantic embedding modules to generate informative initial representations independent of interaction density. Dense sematic embeddings are generated using modality specific neural encoders including transformer based text encoder for descriptive attributes, a recurrent attention network for behavioral interaction sequences and temporal contextual feature encoder for metadata signals. These embeddings are then refined through a normalized graph propagation mechanism that jointly models structural connectivity and semantic similarity, enabling robust higher-order preference learning. Unlike conventional recommenders, the framework preserves attribute semantics during message passing and enables inductive cold-start recommendations, where embeddings for newly introduced users or items are generated directly from attributes without requiring prior interaction edges. Experimental evaluation conducted on publicly available benchmark datasets including MovieLens-1M, Amazon Electronics, Amazon Books, and Amazon Prime Movies and LastFM360 demonstrates consistent performance improvements over Matrix Factorization, content-based models, GraphSAGE and Neural Graph Collaborative Filtering (NGCF). Results show notable gains in ranking accuracy, diversity and robustness under varying sparsity levels. The proposed AE-GNF achieves improved recommendation performance reducing RMSE by 3.5 to 6.2% and improving NDCG@10 by 6-11% compared to graph-based baselines across benchmark datasets. The findings confirm that integrating semantic attribute encoding with graph refinement provides a scalable and effective solution for next-generation recommendation systems operating in sparse and heterogeneous environments.
Most of the existing data center allocation mechanisms contribute either user centric or service provider centric not for both ends but in reality, both have different objectives. For example, the objective of a user is minimization of cost, response time as well as processing time whereas the objective of service provider is to maximize the profit and processing time and minimization of response time, bandwidth, energy consumption and computing overhead with subject to effective resource utilization and load balancing. To address this challenge, this paper introduces a Cost Denigration-Based Data Center Allocation Policy (CD-BDAP) utilizing Particle Swarm Optimization (PSO), which simultaneously considers economic cost, response time, and energy consumption in the selection of data centers. In contrast to conventional PSO-based broker policies, CD-BDAP integrates a workload similarity-aware allocation strategy by calculating a dissimilarity index among user requests, thereby facilitating enhanced consolidation and energy efficiency. A weighted objective function is developed to balance user-centric metrics (cost and response time) with provider-centric metrics (profit and energy consumption), explicitly capturing their trade-offs. The proposed mechanism is assessed utilizing CloudAnalyst, which is constructed on CloudSim. The experimental results indicate that CD-BDAP achieves a reduction in VM cost, a decrease in response time, and an enhancement in energy efficiency, while simultaneously increasing the overall profit for service providers. The findings suggest that the integration of energy-aware cost modeling and workload similarity into PSO-based allocation can enhance both economic and performance efficiency in the selection of cloud data centers. The outcomes of CD-BDAP are compared with the existing PSO-based mechanisms and found enhanced performance.
In the Internet of Things (IoT) environment, a Distributed Denial-of-Service (DDoS) attack in the network causes poor performance and resource-limited issues to users. Existing systems do not provide real-time adaptability, leading to delayed mitigation. Also, centralized storage systems suffer from breaches and tampering. To tackle these issues, a secure and intelligent IoT DDoS detection and mitigation framework is presented that utilizes hybrid encryption, blockchain storage, ensemble deep learning (DL), and reinforcement learning (RL) to improve the accuracy, security, and efficiency of IoT networks against several cyber-attacks. The developed technique collects data from a dataset and pre-processes it for handling missing values and normalizes it for further analysis. Secondly, a hybrid encryption method combining Homomorphic Encryption (HE) and ChaCha20 is adopted for data encryption with optimal key selection using Dingo Optimizer (DOX). Then, the encrypted data is securely stored in blockchain through off-chain storage and on-chain hash storage to ensure data integrity and tamper-proof security. DDoS attack detection is performed using an ensemble model called TriGuard-Net that combines AlexNet, LSTM, and PSPNet, with optimizing hyperparameters using Fire Hawks Optimizer (FHO). Finally, an RL-based mitigation system using Deep Q-Network (DQN) helps in real-time attack mitigation and enhances IoT security. Experimental results reveal that the presented model offers superior performance by achieving an accuracy of 99%, a kappa score of 98%, an R2 Score of 97%, an MCC of 98%, a Jaccard Score of 98%, and a Hamming Loss of 0.006, thereby outperforming other current models.
