In recent years, the growth and popularity of electric vehicles (EVs) has soared owing to the facilitation of zero-emission carbon for people commuting on the road, preserving the environment from air pollution and hazardous gases. However, uncertain EV energy demands and their dynamic arrival times impact the ancillary operations and stability of the charging station (CS). Thus, it becomes a challenging task to schedule EVs for charging with their dynamic charging prices, traveling time, and waiting time efficiently and optimally. Thus, we propose an optimal EV selection scheme for trustworthy charging by implementing the hybrid game theory. The hybrid game theory is bifurcated into stage 1 and stage 2, in which stage 1 includes a coalition game to generate EV clusters or coalitions based on the parameters of state-of-charge (SoC), energy demand, and penalty factor. Then, the trust values are determined to select the EV pair fairly. Furthermore, stage 2 highlights the zero-sum game theory, which aims to optimize the payoff at saddle point and formulate strategies for EV pair (generated in stage 1), ensuring the optimal EV selection for trustworthy charging. Moreover, we have utilized the blockchain network to secure the EV optimal payoff by implementing smart contract in Remix Integrated Development Environment (IDE). The hybrid game theory ensures the optimal and efficient EV selection using coalition game to select EV pair then apply zero-sum game to optimize the payoff at saddle point condition. Next, we implement the hybrid game theory in Python 3.9 to simulate the results with the help of various factors such as trust value comparison, profit comparison based on strategies, convergence comparison, and profit comparison with the traditional approach.

The malicious node attacks on Proof-of-StakeProof-of-Stake (PoS) blockchainBlockchain are at risk to the securitySecurity and the legitimacy of autonomous vehicleAutonomous vehicles systems in supply chainSupply chain management. This attack compromised the integrity of important data exchanges and various operations. In this work, we have put forth a framework for detecting rogue nodes in PoS blockchainsBlockchain that is based on Explainable AIExplainable AI (XAI) and specifically tailored for applications involving autonomous vehiclesAutonomous vehicles (AVs). This framework implies several machine learningMachine learning models, which include models such as K-Nearest Neighbors (KNN), Gradient Boosting Machine (GBM), Decision Tree (DT) and Naive Bayes (NB) to analyze node behavior and to spot and identify anomalies such as delay, conspiracy, and interference. KNN obtained 0.8747, the highest accuracy score. We have investigated significant properties such as block height, coin age, and transaction timestamps using XAI techniques, including SHAPSHAP and LIMELIME; it increases openness and confidence in the machine learningMachine learning model. Metrics including precision, recall, F1-score, and ROC curves help us to calculate the performancePerformance so it promises strong evidence of abnormal and malicious activity. This method provides a reliable and trustworthy way to protect our vehicle operations in supply chainsSupply chain so guaranteeing a safe and reliable blockchainBlockchain data exchanges.

This paper presents a secure and noise-resilient communication scheme which has been designed for proximity-based device-to-device networks. A set of 50 original signals having the power range between 23-25 decibels were generated and used as a set of real signals. A generative adversarial network (GAN) was then used to synthesize 500 deep-fake signals that were used to emulate the statistical structure of the original 50 signals. The signals were passed through a noisy channel which had the noise of additive white Gaussian noise (AWGN). At the receiving end, a supervised denoising neural network was trained using the clean and noisy signals to recover the original signal. The receiver was able to regenerate the original signal without any error and the correlation of the original signal and the signals regenerated by the receiver was high. In contrast, an eavesdropper (C) was trained only on one intercepted signal without access to clean references and the GAN network coefficients. The eavesdropper failed to reconstruct the meaningful signals and had very low correlation with the original signal. The strong denoising ability of the receiver and the communication security is validated by the evaluation metrics, signal-to-noise ratio, correlation coefficient, and Shannon capacity.

Backscatter communication technology is used for optimizing energy efficiency in wireless communication networks. It leverages existing signals in the environment rather than generating new ones, allowing devices to communicate with minimal power consumption. Energy efficiency of wireless networks is a key goal in futuristic communication systems, the growth of IoT, and 5G beyond. Various methods have been proposed to enhance energy efficiency in wireless networks, focusing on reducing energy consumption while maintaining network performance. In this paper, the suggested technique combines energy harvesting with non-orthogonal multiple access technology to achieve optimal energy efficiency in wireless networks. The proposed technique is compared with the orthogonal multiple access technology of the network. Results show the efficacy of the proposed technique compared to the state-of-the-art approaches in terms of energy efficiency, distance of backscatter, and power splitting ratio.

With the trending field of Internet of Things (IoT) and smart homes (SHs), there are cybersecurity risks associated with it. Hence there is a need of effective and transparent intrusion detection system (IDS). Existing IDS models often lack explainability and robust and secure communication, thus creating a research gap. We experimented using the RT-IoT2022 dataset, which gives IoT network attack data. We apply information gain-based feature importance to filter out relevant features and apply machine learning models. Then we integrate LIME-based Explainable AI (XAI) to add an explainability. Classified data is passed through Quantum Channel for secure communication. We achieved a significant accuracy with XGBoost of 99.7

With the development in the healthcare industry, the main problem that remains is the absence of proper infrastructure in rural areas to support it. This can be fought by using telesurgery, that is surgeons can perform operations on patients virtually with the aid of robotic systems and high-speed Internet connection. However, since the surgeries are performed virtually, other types of cyber threats may emerge. Hence, in this paper, we propose a TinyML framework to accurately classify a data packet as arrived due to a cyber-attack or just a normal data packet. Our proposed TinyML takes a considerably smaller space than other models and frameworks that are employed for this purpose. We trained 4 models: our proposed TinyML model, LSTM, GRU, and 1D CNN, and also compared these models based on their accuracy, loss, and memory consumption. When evaluating the performance of TinyML with the other competing models based on the three performance metrics, TinyML was seen to be superior in the - accuracy, loss, and memory size metrics. Regarding testing accuracy, TinyML had the highest accuracy with NAdam Optimizer at 98.002% while LSTM, GRU, and 1D CNN had 97.5411%, 97.9707% and 96.88% respectively. It also had the minimum model size of 0.14 MB when used with RMSprop optimizer Other models including LSTM, GRU, and 1D CNN had a model size of 0.48 MB, 0.39 MB, and 0.84 MB respectively, all used RMSprop optimizer.

Maternal healthMaternal health issues now continue to be a major challenge for the world over. Early risk predictionPrediction willRisk prediction help in timely management and in improving the outcomes. This study proposes a pioneer smart contract system providing machine learningMachine learning (ML) integration for predicting maternal healthMaternal health risk in a data-secure way. Under this framework, patient data such as age, blood pressure, blood sugar, body temperature, and heart rate can be classified into different risk levels with the help of the conventional ML algorithms. By providing dataSecurity securityData security, blockchainBlockchain technology with the AES encryption guarantees secure communication and storage, thus contributing to transparency and integrity of the data. Therefore, this study takes an extra step toward moving healthcare technologyHealthcare technology forward by merging robust ML predictive capabilities with cutting-edge decentralized securitySecurity.

Autonomous vehicles have become increasingly popular from the last few years because of their numerous benefits, such as minimum travel time, enhanced safety, and improved fuel economy. Many vehicle manufacturing companies and tech giants are working on this technology to make fully autonomous vehicles or strengthen their existing driver-less vehicles. These vehicles use complex, advanced, and sophisticated hardware technologies. However, the software is an equally important feature because it must operate all functions seamlessly in sync with other vehicle components. The software must analyze large volumes of data to make quick real-time decisions, so any vulnerabilities or bugs can be a severe problem to the vehicle and the passengers riding in it. Many researchers have proposed various software defect prediction schemes for different projects and applications, but most of them have focussed on specific software issues and excluded others. Thus, their methods cannot be applied to the software of autonomous vehicles. In this paper, we propose an improved Artificial Neural Network (ANN) model, called Dropout-Artificial Neural Network (D-ANN), to solve this problem of defect prediction in autonomous vehicles. This inclusive model can consider all the parameters simultaneously for effective bug prediction. The proposed model can be used for the software of any autonomous vehicles, and it is trained and evaluated using standard methods. The results obtained show that the proposed model predicts software defects with higher accuracy than other models.

Anomaly detection in electronic health records (EHRs) is a critical task due to the substantial amount of large and unstructured data. Existing machine learning (ML) approaches often fail to capture the complex relationships present in clinical text. Natural language processing (NLP) task handling has been notably improved using transformer-based models, such as BERT, BioBERT, and ClinicalBERT, in recent breakthroughs. In this study, we introduce a Graph Risk Extraction (GREx) framework that combines pretrained transformers with graph-based relational modeling to detect threats in healthcare. The proposed approach derives an entity graph from clinical notes using ontology relationships and co-occurrence and applies a graph convolutional network to model multi-hop dependencies among entities. Evaluations on datasets MIMIC-III and i2b2 show GREx outperforms the ClinicalBERT baselines. The proposed framework achieves an accuracy of up to 86%. The applications such as work on sepsis detection, adverse drug event monitoring, and hospital readmission prediction demonstrates the practical value of the framework.

Smart grid (SG) has revolutionized the traditional energy sector by reducing its regular electricity costs, helping the environment, and saving finances. However, SG is prone to many security flaws (e.g., data manipulation, denial-of-service, and man-in-the-middle attacks), particularly when relaying smart grid data over the public Internet, using conventional routing techniques. To tackle the aforementioned security problems, we propose an artificial intelligence and onion routing network-based secure data exchange architecture in the SG environment. Onion routing uses anonymous, encrypted routing mechanisms to address prevailing network-related attacks from the smart grid ecosystem. To reduce computational overhead in the onion routing network, we employed machine learning algorithms to classify as malicious or non-malicious before exchanging it between components. We further enhance the security of conventional onion routing networks by incorporating verifying tokens that verify the legitimacy of onion routers. Routers are authorized to forward SG data upon successful validation, ensuring that only trusted nodes participate in the network. To safeguard verifying tokens from data integrity attacks, they are securely stored in the immutable ledger of a blockchain, preventing unauthorized modifications or tampering. Additionally, to mitigate latency issues, we leverage high-speed, low-latency capabilities of 5G networks, significantly improving throughput, reliability, and real-time data transmission. Lastly, the proposed framework is evaluated by considering different evaluation metrics, such as artificial intelligence statistical measures (accuracy (98.86%)), compromisation rate (> 25%), onion router selection, data compromisation and an anonymity rate, and throughput (28.23 Mbps).

The desirability of both the industry and academia is stretched by the expanded performance in the field of fifth generation (5G) communication technologies, consumer electronics, big data analytics, agriculture, and healthcare. Several challenges such as security challenge, design challenge, deployment challenge, and so forth exist in these fields. An apparent and obvious problem is the bounded lifetime of the network. The justification is that massive devices are generally powered by batteries with finite capacities. A new solution has been introduced to address this challenge. That is a very promising technology named backscatter communications (BackCom) with two special features as low-power and low-complexity. In BackCom, the concept of passive reflection as well as modulation of an incoming radio frequency wave is used. In 5G and beyond fifth generation (B5G) network, nonorthogonal multiple access (NOMA) technique is highly employed to increase the capacity of communication networks. In case of multi-cell NOMA networks, the union of these two techniques gives us new outcomes. This new technology gives rise to several new challenges in dense wireless network. In this article, capacity of BackCom network with two transmission techniques is calculated with the consideration of signal-to-noise ratio, reflection coefficient, and path loss exponent individually. A comparison has been provided between these two techniques to show the improved performance of NOMA BackCom with OMA.

Cancer is the second leading cause of death worldwide almost 10 million people die every year due to cancer. Lung cancer is one of the most common cancer found in both women and men, and to increase the survival rate, early diagnosis is the key. But due to an increase in population, it has become very difficult for radiologists to view every Computed Tomography scan and make a correct diagnosis. This created an opportunity for the penetration of Artificial Intelligence into the medical sector. Convolutional Neural Networks are a specific type of artificial neural network that can understand some specific patterns in an image easily. While this model provides a higher range of accuracy, it also has high energy consumption, higher computational cost, and higher memory space. To overcome these challenges, TinyML was introduced in 2019. In this paper, we have trained four models: MobileNet-V1, MobileNet-V2, EfficientNetB0, and CNN with transfer learning of MobileNet-V3 on a lung cancer dataset from Kaggle. After training, we quantized these models with the help of Post Training Quantization(PTQ) methods. Out of which the best performing model was MobileNet-V2, whose pre-quantization accuracy was 97.78%, and after quantization, its accuracy increased to 97.92%. The results of the study have the potential to provide radiologists with a tool that they can trust for efficient and reliable diagnosis of lung cancer in its early stages.

Unmanned Aerial Vehicles (UAVs) have gained widespread attention in diverse applications like military, medical, aerial surveillance and many more. Presently, the problem of limited bandwidth and geographic factors has raised the need for effective and timely data transfer. Training UAVs with reinforcement learning-based algorithms facilitates autonomous decision-making capabilities. In this paper, we proposed an intelligent system for the optimal UAV selection process by evaluating the continuous performance of each UAV. The analyzing factors are based on the real-world factors affecting the quality of signals, such as noise interference, relative motion between source and wave, and transmission power. Based on the systematic conditions observed, the system provides efficient rewards. To promote the selection of the optimal UAV and enhance the learning process, the state information of the UAV is fed into a deep neural network (DQN), which predicts the 'Q-values'. Our system implements a deep Q-learning algorithm, which enhances the agent's performance by systematically learning from its experience. The model operates accurately by selecting the most reliable UAV, thus, enhancing the throughput by optimal power allocation. It outperforms other conventional models in terms of timely data delivery and energy utilization. The system adapts various complex patterns by analyzing the historical and present scenarios. Empowered by this intelligent system, time-critical decision-making can be achieved with minimal energy consumption.

Unmanned Aerial Vehicles (UAVs) are used in various civilian applications to collect data. Security is a major issue in UAV networks. To securely transmit the data from the UAVs to the base station, the data must be sent in encrypted form. Cryptography techniques are used for encryption and decryption. Classical cryptography schemes do not provide protection against quantum attacks. To secure the data from quantum threats we use post quantum cryptography(PQC) techniques. In this paper, a post-quantum secure intrusion detection model named the PQ-TabNet model is proposed to address these challenges. The proposed model integrates N-th Degree Truncated Polynomial Ring unit (NTRU) lattice-based encryption with a transformer-based deep learning model to ensure confidentiality and resilience against both classical and quantum attacks. The proposed model was trained on the CICIDS2018 data set, where PQ-TabNet achieved better performance than the Kyber-based configuration in terms of energy efficiency and latency. Experimental analysis demonstrated that NTRU provided faster encryption and reduced energy consumption while maintaining strong quantum resistance.

The growing digitalization of Computer Numerical Control (CNC) machines in the current technological era has brought forth a new set of cybersecurity risks, particularly False Data Injection Attacks (FDIAs). These attacks work by manipulating sensor data and control parameters to disrupt the manufacturing process, leading to a variety of issues like compromised product quality, increased maintenance costs, and production downtime. Because of the subtle nature of these attacks, traditional cybersecurity measures often fail to detect such attacks. This research aims to use machine learning models like Decision Trees, XGBoost, etc. to detect FDIAs. Experimental results show that Random Forest achieved the highest performance with an accuracy of 99.68%, making it a promising solution to the previously mentioned problem. Additionally, SHAP (SHapley Additive Explanations) is employed not only for feature selection but also to provide interpretable insights into model predictions by identifying the most influential system and network-level parameters contributing to FDIA detection. Such explainability is crucial in safety-critical Industry 4.0 environments. It enables CNC operators and industrial security engineers to validate alerts, perform root-cause analysis, and make informed operational decisions. Future work can focus on real-time deployment, adaptive learning, and edge-based implementation to further enhance the practical applicability of the proposed framework.

In this paper, we propose FedChain, a secure and decentralized blockchain-enabled federated learning (FL) framework for resource allocation (RA) in device-to-device (D2D) communication networks. Traditional D2D RA methods are based on centralized processing, which introduces privacy risks, single-point failures, and limited scalability. FL decentralizes model training, but remains vulnerable to model tampering and insecure parameter exchange between clients and the server. To overcome these challenges, FedChain integrates a permissioned Hyperledger Fabric (HF) blockchain to provide traceable, auditable, and tamper-proof model update transactions. Additionally, the interplanetary file system (IPFS) is employed to store model parameters off-chain, while only their cryptographic content identifiers (CIDs) are recorded on-chain to minimize blockchain storage overhead. The framework applies federated averaging (FedAvg) for collaborative model aggregation without exposing raw data. An experimental evaluation compared the results of 5 models, where Fedchain achieves >95% global accuracy using Random Forest (RF) and eXtreme Gradient Boosting (XGBoost) algorithms. It helps improve security, tamper-resistance, model consistency, and communication efficiency. These results indicate that FedChain offers a scalable and secure foundation for decentralized RA in future 6G-enabled D2D networks.

The Industrial Internet of Things (IIoT) is transforming the industrial landscape by enabling real-time data exchange and automation, reducing the need for human intervention and enhancing accuracy beyond manual labor. However, this interconnectedness creates a vast assortment of security vulnerabilities, ranging from device-level attacks to a complex nexus of network breaches. This paper introduces a comprehensive security framework utilizing Machine Learning (ML) for smart threat identification and Explainable AI (XAI) to improve transparency and build trust in decision-making processes. To effectively tackle the dynamic and complex challenges of IIoT environments, we introduce Large Language Models (LLMs) as an innovative and critical layer within our security pipeline. LLMs are utilized for contextual threat evaluation and support in real-time incident response. We have also demonstrated that the incorporation of ML, XAI, and LLM technologies not only boosts detection precision and speeds up response times but also improves human comprehension and operational effectiveness. The XG Boost model performed the best amongst all the ML models, achieving accuracy. We further used LIME and SHAP as the XAI tools to provide the reasoning behind the model’s prediction. The Judge and Lawyers models are implemented as a part of the LLM Layer to provide precise diagnostic information as a part of security threat detection.

Disease management and prediction have greatly improved with the growing use of Machine Learning (ML) techniques in healthcare. In order to avoid serious consequences, Diabetes Mellitus, a common chronic disorder, requires a precise and timely diagnosis. In addition to providing an analysis of current ML based techniques for prediction of diabetes, this work includes an empirical case study utilizing the PIMA Indian Diabetes Dataset, which consists of 768 patient data classified into eight characteristics (pregnancy, glucose, blood pressure, skin thickness, insulin, BMI, and function of the diabetes pedigree). Four classifiers were used: nearest neighbors (KNN), random forest, decision tree, and support vector machine (SVM) and their diagnostic precision was assessed. With the best accuracy of 76.62%, the SVM was followed by KNN (66. 23%), Random Forest (72. 07%) and Decision Tree (75.97%). Finding successful models for early diabetes prediction by combining empirical validation with a literature-based comparison analysis is what makes this work novel. Additional clinical and lifestyle parameters should be included in future research to improve model generality and practicality.

Human activity recognition (HAR) is vital to support day-to-day human activities and improve the interactions among users and external objects. HAR provides vital information about user personality and physiological traits, which can be analyzed through effective Artificial Intelligence (AI) based models. Different applications in computer vision and Machine Learning (ML), like video surveillance, healthcare, gesture recognition, person identification, and human-computer interaction, utilize the key principles of HAR. Recent surveys have centrally focused on the role of AI in HAR, where the computational requirements (both centralized and decentralized) are not discussed in detail. Owing to the gap, the article presents a detailed review of the fusion of AI and the cloud to address the dual objectives of model learning, resource requirements, and optimization. Based on research questions, we present a reference architecture that integrates AI and cloud for HAR in surveillance systems. A solution taxonomy is presented for diverse industrial verticals, and key challenges and future opportunities of HAR are discussed. A case study of Deep Learning (DL) based HAR is proposed for smart healthcare, where networking and security parameters are considered. The survey intends to assist industry and academia in the design of novel cloud and AI-based HAR systems with optimal control and resource management in various applicative verticals.