Mobile Ad hoc Networks (MANETs) are a rapidly developing technology, making their security a major concern. In order to enable trust-aware and attack-resilient routing, the goal of this research is to develop an intrusion detection system (S-IDS) based on deep learning (DL). By leveraging trust values of the nodes using Osprey Optimization Algorithm (OOA), the presented Dilated Convolutional Multi-Relational Graph Attention Network (DConMRG-Net) based Intrusion Detection System (IDS) identifies potential intruders, ensuring that the paths generated within the MANET are reliable and resilient. Additionally, a novel optimization algorithm namely, Hybrid Adaptive Genghis Khan Shark Gold Rush Optimization (HAGKS-GRO) Algorithm is introduced by combining the Adaptive Genghis Khan Shark Optimization (AGKSO) Algorithm with Gold Rush Optimization (GRO) Algorithm for optimal path selection. Two situations are examined: one in which there is no attack and the other in which there is an attack. Relevant performance metrics are evaluated in connection with these scenarios, including throughput, packet delivery ratio, attack identification rate, accuracy, error rate and computation time. Evaluation results demonstrate significant improvements, with a maximum detection rate of 99% with a minimum computational time of 55ms for with attack case and 51ms for without attack case. The proposed model surpasses the state-of-the-art attack detection techniques and achieves high efficiency, according to the simulation results.

Cloud computing has transformed data management for businesses and individuals alike by making systems more scalable and economically viable. However, the distributed architecture in cloud computing inherently makes it vulnerable to highly sophisticated cyber threats, such as DDoS attacks, ransomware, cryptojacking, and many others that could compromise data confidentiality, integrity, and availability. Traditional intrusion detection systems face limitations such as high false-negative rates and low adaptability to emerging threats. This manuscript proposes a blockchain-integrated discrete Hopfield neural network and edge attention networks with duck swarm optimization (Hop-MEA-Duck) to enhance cloud privacy and develop a robust, privacy-preserving intrusion detection framework for cloud environments. The framework assumes a two-level privacy mechanism. The former provides privacy through the Adaptive Blockchain Sharding Protocol with Hybrid Consensus (Hyb-BCSP), which enhances data security and scalability. The second tier comprises preprocessing using the Adaptive Self-Guided Loop Filter (ASGLF) and feature selection via the Pufferfish Optimization Algorithm (POA). The Discrete Hopfield Neural Network (DHNN) with Multilayer Edge Attention Network (MEAN) is used to classify normal and abnormal behaviors, and the Duck Swarm Algorithm (DSA) is used at the expense of hyperparameter tuning. The results of the experiments indicate that the proposed framework is practical, achieving 97.8% detection accuracy, 96.5% precision, and a significant reduction in the false-negative rate to 2.4%. Also, the preprocessing phase increased the relevance of the data by 85%, whereas the rate of information immutability in the blockchain was 99.2%. To sum up, the suggested framework can provide a privacy-preserving, scalable, and adaptable solution for detecting and removing cyber threats in the context of cloud computing. It can fill significant gaps in current intrusion detection approaches.

The rapid expansion of the Internet of Things (IoT) has enabled real-time patient monitoring through medical sensors, but redundant readings significantly increase storage, transmission, and energy costs. To address this challenge, we propose the Energy-Schoof’s Cryptography-based Similarity-Navigated Graph Neural Network with Human Memory Optimization (ESC-SNGNNet-HMO) for healthcare data aggregation. The framework efficiently manages data chunks by combining deduplication, energy-aware processing, and secure transmission. At the fog layer, a Similarity-Navigated Graph Neural Network (SNGNN) identifies duplicate records, with hyperparameters optimized through Human Memory Optimization (HMO) to enhance accuracy. Deduplicated data is then securely transferred to the cloud using Schoof’s Dynamic Elliptic Curve Cryptography (SDECC). Experimental evaluation demonstrates that ESC-SNGNNet-HMO maintains a throughput of 230 KB/s even with 8% packet loss, reduces storage to as little as 14 bytes, and eliminates up to 99% of duplicate data in low-node scenarios. Overall, the system provides an energy-efficient and cyber-secure solution for redundancy management in IoT-based healthcare applications.

An electroencephalogram (EEG) is a critical diagnostic tool that monitors brain activity and detects epileptic seizures. The EEG signals, however, are complex and unwieldy, making it challenging to create an automated seizure-detection system. Additionally, sensitive healthcare data in clinical systems should be handled securely. The given paper describes an EEG-based seizure detection using blockchain security and transformer graph capsule networks (PT-GG-CapsNet-Blockchain). This model integrates a superior signal processing system, machine learning, and blockchain to attain powerful seizure tracking and data protection. The structure takes EEG data of the CHB-MIT dataset, which consists of multichannel records of seizures in children. Preprocessing was performed using the Adjoint Bilateral Filter (ABF), which removes noise while preserving critical signal features. Using the Clifford Fourier Mellin Transform (CFAT), spatio-temporal features are extracted and thus capture the complex dynamics in EEG signals. Classification will be performed by a Pure Transformer-Based Gated Graph Attention Capsule Network (PT-GG-CapsNet), leveraging transformers' temporal analysis capabilities and graph capsule networks' modeling of spatial relationships. The Artificial Hummingbird Algorithm (AHA) will then be used for hyperparameter optimization, ensuring an optimal model. To achieve data security and transparency, the framework incorporates blockchain technology to allow decentralized storage of EEG data. This guarantees integrity, immutability, and effective coexistence among heterogeneous medical networks. The classification accuracy, precision, recall, and F1-score of the proposed model are 99.7%, 99.5%, 99.2%, and 99.3%%, respectively, whichbeats baseline methods. This framework will offer real-world clinical uses of blockchain that can operate in a heterogeneous medical setting, thanks to its low latency and interoperability.

As the cloud computing and mass data sharing develop, data integrity and privacy has become an imperative issue. Conventional remote data auditing techniques tend to reveal sensitive data or they have high computational cost. In order to overcome these shortcomings, the Fully Homomorphic Encryption enhanced Remote Method Invocation (FHEbRMI) mechanism that includes a combination of the Modified Least Squares (MLS) optimization model and the proposed cloud auditing security and efficiency are proposed in this paper. The suggested system provides an encrypted data auditing system, which involves RMI-based communication, to enable the client, server, and third-party auditor to perform their verification functions remotely without the disclosure of the plaintext data. An actual execution of the suggested structure is introduced, such as secure key generation, trapdoor-based dimensionality reduction, ciphertext multiplication, and optimized homomorphic functions. Moreover, the RMI interface provides a smooth communication among the distributed nodes and increases the scalability and minimizes transmission delays. A comparative study with the recent homomorphic-based auditing schemes like blockchain-assisted, certificateless and lattice-based FHE model reveals that the proposed FHEbRMI-MLS model has better performance in terms of encryption/decryption latency, computational cost, and encryption overhead. The experimental performance is indicative of an average 37 and 42 factor in speed of encryption and enhancement of computational efficiency respectively with respect to the traditional FHE models. This paper presents a viable, privacy-friendly auditing framework of clouds which guarantees the end-to-end encrypted verification without sacrificing the efficiency.

The rapid evolution of the Internet of Things (IoT), supported by the convergence of cloud, edge and mist computing layers, opens new avenues for delivering reliable and responsive services to distributed smart devices. However, ensuring efficient and adaptive service replication in such resource-constrained and dynamically changing IoT environments remains a significant challenge. To tackle this, we introduce Elastic Context-Aware Replication (ECAR), an intelligent replication strategy tailored for IoT systems. ECAR dynamically redistributes services across the IoT continuum by leveraging both physical and logical contextual information. Unlike traditional replication schemes, ECAR continuously adapts to real-time workload fluctuations and network conditions, ensuring low latency and efficient resource usage. ECAR’s effectiveness is demonstrated through a comparative evaluation involving diverse IoT deployment scenarios, including Cloud-Intensive Replication (CIR), Cloud-Edge-Intensive Replication (CEIR) and Cloud-Edge-Access-Intensive Replication (CEAIR), alongside two existing replication strategies, Group-Delay-Aware Replication (GDAR) and Combined Context-Aware Replication (CCA). The evaluation shows ECAR achieving up to 18% reduction in service drop rates, 82% improvement in allocation efficiency and 82% better resource utilization. These results underline ECAR’s effectiveness in supporting scalable, reliable and latency-aware service delivery for IoT deployments.

Neuromorphic computing is a paradigm based on the computational mechanisms of the human brain and has received considerable attention as a real-time technique with low energy requirements. Present systems, however, are limited in their ability to scale traditional processors to a neuromorphic architecture, leading to issues with latency, power consumption, and smooth data flow. To address these problems, this paper proposes the ACORISC-VbSNN framework, comprising a modular RISC-V architecture, Spiking Neural Networks (SNNs), and Ant Colony Optimization (ACO). The system uses a shared-memory architecture to maximize communication between traditional and neuromorphic processors, ensuring data is managed effectively. The postulated framework processes the sensory data by pre-processing and encoding them using rate coding, and dynamically optimizing memory access. SNNs are also used to process spike trains in real-time, whereas ACO is used to determine the best data paths to minimize bottlenecks. Experimental analysis shows that the system performs better, with ultra-low power consumption of 0.0095 mW, very low latency of 0.000544 seconds, and 99.2 percent accuracy. These findings indicate that the ACORISC-VbSNN model has the potential to advance the field of bio-inspired computing, providing a highly accurate, energy-efficient, and low-latency system for real-world use.

In recent years, AUV-assisted underwater sensor networks (AUV-UWSNs) have gained significant attention due to their ability to monitor and collect crucial data from the oceanic environment. However, efficiency of these networks has hindered certain issues which include energy consumption problems, low communication range and rational routing/ clustering problems. Two of the major challenges affecting other aspects of AUV-UWSNs are the lack of a proper way to choose the right cluster heads and guarantee sound data transmission. This paper proposes a high-efficiency clustered routing technique that tackles these problems by utilizing the Spider Wasp Optimizer (SWO) for optimal cluster head selection and the Deep Fuzzy Gorilla Troops Self-Guided Generative Adversarial Clustering Network (DFGTS-2GACN) for efficient data transmission. The SWO enhances the choice of cluster heads with respect to energy consumption and the DFGTS-2GACN applies deep learning and fuzzy logic control to facilitate data forwarding in the network with enhanced performance. Experimental performance shows 99% improvement compared with the existing methods. In this paper, an effective method to consider for increasing the data transfer rate while decreasing the energy consumption plus increasing the life cycle of the underwater network facilitated by the AUV in the underwater sensor networks is discussed, thus providing a viable solution to the problems of underwater communication.

Spam emails continue to be a major issue in digital communication, hurting productivity and jeopardizing security. This paper presents a real-time spam email filtering method that combines ensemble learning techniques (stacking and majority voting) with machine learning classifiers such as Logistic Regression (LR), Support Vector Machine (SVM), Stochastic Gradient Descent (SGD), and Decision Trees (DT). The objective is to improve the accuracy and reliability of spam detection systems. The proposed models were tested against Kaggle’s Email Spam Dataset on crucial performance parameters such as accuracy, precision, recall, F-measure, false positive rate (FPR), and false negative rate (FNR). The results show that, while the stacking ensemble is competitive, the majority voting ensemble consistently achieved good results. It delivers improved accuracy of 98.59%, precision of 98.59%, recall of 98.59%, and F-measure of 98.59% while keeping lower FPR of 0.019 and FNR of 0.009, making it the most dependable option for real-world use. This study demonstrates the efficiency of ensemble learning, particularly majority voting, in building scalable and practical solutions for real-time spam email filtering. The findings lay a solid foundation for enhancing email security mechanisms and user experience.

In big cloud data centers, the best physical machine (PM) is selected using the Virtual Machine Placement (VMP) process. To address this issue, a number of approaches have been proposed. Nevertheless, the existing solutions only take into account a small number of resource categories, which leads to an uneven load and ultimately, the activation of superfluous physical computers within the data center. The aim of this research is to maximize resource usage while lowering power use and carbon footprints by integrating a Hybrid Swarm Bipolar with Walk-Spread Algorithm with a unique Deformable Graph Convolutional Network (DGCN-SB-WSA)-aware virtual machine placement architecture. To ensGoogleent VM scheduling and management, real-world cloud workloads are analyzed using the Google Cluster Dataset (GCD). The Deformable Graph Convolutional Network (DGCN) dynamically models cloud infrastructure as a graph that captures intricate relationships among PMs and VMs, enabling adaptive placement choices. The Hybrid Swarm Bipolar with Walk-Spread Algorithm (SB-WSA) then uses a dual-phase search approach to balance local exploitation with global exploration, minimizing premature convergence and increasing performance while optimizing Virtual machine (VM) allocation. Comparing the proposed method to conventional VM placement techniques, experimental results show that it dramatically 27 KW lowers power consumption, improves 98% resource usage and 180 kg CO₂ decreases carbon emissions.

Internet of Things (IoT) has revolutionized mobile healthcare applications, increases diagnosis speed and accuracy. Disease prediction systems (DPS) improve healthcare quality, but they raise privacy concerns due to sensitive data. These concerns include illegal sharing, misuse, and exposure of sensitive information. However, developing new techniques does not provide improved protection from attackers and fraudsters. This paper suggests an effective privacy-preserving strategy for patient healthcare data from IoT devices in order to anticipate diseases in the contemporary medical field. Heart failure prediction health data is utilized as an input in this proposed approach. Initially, elastic net (EN) is used to reduce the dimensionality of the input raw dataset. Hyper parameter in EN is optimally selected using the Zebra Optimization Algorithm (ZOA). After dimensionality reduction, data is encrypted using the ElGamal technique. During the encryption procedure, the True Random Number Generator-Pseudo Random Number Generator (TRNG-PRNG) encryption method is used to generate the secret key. These encrypted data is securely stored in cloud. Finally, Attention Mechanism based Bi-directional Gated Recurrent Unit (AT-BiGRU) technique is employed to predict heart disease. The ElGamal-TRPRNG method strengths privacy and security by achieving encryption and decryption times of 0.30 sec and 0.12 sec, respectively. The suggested model is evaluated and contrasted with current methods using encryption data performance measures. Model achieved 93.47% accuracy, 6.53% error, and 93.45% precision in encrypted data performance metrics. Therefore, this suggested method is the most effective way to effectively safeguard health care records.

Cognitive image authentication systems implement specialized behavioral and cognitive parameters for accurate and secure user verification. This paper presents a dual-Face cognitive Cloud image authentication framework for improved accuracy and secured cognitive biometric features. The current work addresses the endpoint weaknesses of malicious hacking and biometric systems. The MobileNetV2 model along with XOR segmentation and encryption safeguard the system’s real-time authentication. Essential features include optimizing facetime image and video over the handheld swiping devices to the cloud and the retention of authentication accuracy regardless of ageing, accessories, and vanity. Additionally, the performance of MobileNetV2 cognitive authentication and the rest of the systems were tested and analyzed. An experimental result comparison with the rest of the systems showed that the proposed MobileNetV2 based model achieved over 95.3% accuracy and a 91.11% F1 score for real-time and large-scale datasets respectively, outperforming the rest of the models including CNN, ResNet50, VGG19, hence able to work as a satisfactory cloud-based authentication. The proposed system showed response time for the XOR encryption method of 55ms to 75ms, which was also 63-68% faster than the proposed system using AES encryption. Cross-architecture performance evaluation with several deep learning systems confirmed the performance of MobileNetV2 as the best available for cloud scalable authentication. The proposed system offers a major cognitive biometrics advancement by providing a lightweight, efficient and robust authentication framework.