Accurately classifying petrol and diesel fuel using an image processing method is crucial for fuel-related industries such as petrol pumps, refineries, and fuel storage facilities. However, distinguishing between these fuels using traditional methods can be challenging due to their similar visual characteristics. This article introduces a novel multi-scale and multi-level modified histogram of oriented gradients (MHOG) feature descriptors for robust classification of fuel images. Our proposed method involves extracting distinctive features from the images using the novel multi-scale and multi-level MHOG feature descriptor. These features are then utilized to train a range of machine learning classifiers with different hyperparameter settings for an ablation study. To the best of our knowledge, this is the first ablation study for this fuel classification application. To evaluate the effectiveness of our approach, we conduct experiments on a carefully labeled dataset consisting of petrol and diesel fuel images. The results demonstrate the high accuracy of our proposed method, achieving a classification accuracy of 98% using the light gradient boosting machine (LGBM). Furthermore, our method surpasses existing state-of-the-art techniques for fuel image classification. With its superior performance, this approach holds great potential for efficient and effective fuel classification in diverse fuel-related industries.

Data scarcity is a crucial concern in the traditional approach of training deep learning and one of the bottlenecks that limit its growth. Recently, Federated Learning (FL) has become a suitable approach for providing data privacy and is emerging as a solution for data scarcity. However, FL has opened up a new issue, i.e., model privacy and security. In vanilla FL, each participant receives the updated global model in every training round. Hence, if a model trainer wants to keep the updated global model private from participants, there is limited scope to protect the model access. There needs to be more literature on preserving the global model, and possible solutions like differential privacy, cryptography, or subnetworks are insufficient. In the proposed work, we have introduced the privacy issues in the global model and provided experimental results to demonstrate global model leaks, i.e., each participant has a model with equivalent accuracy to the global model in the subnetwork-based FL approach.

Federated learning has emerged as a highly promising approach for training machine learning models across a decentralized network of clients, with a key focus on maintaining the privacy of data. Nevertheless, the management of system heterogeneity and the handling of time-varying interests continue to pose hurdles for conventional federated learning methodologies. This work presents temporal-based adaptive clustered federated learning as a viable solution to the difficulties mentioned above. The evaluation of clusterability is conducted by calculating the Silhouette score following each iteration of federated training. The process of model aggregation is performed at the cluster level, resulting in enhanced convergence efficiency and improved accuracy of predictions. The inclusion of temporal-based adaptiveness in clustered federated learning for time-varying environments enables the system to dynamically modify cluster configurations in response to clients joining or leaving the network. The experimental results on a real-world dataset of an electric vehicle charging station network illustrate the efficacy of the suggested approach in terms of model correctness, convergence, and adaptability. The temporal-based adaptive clustered federated learning framework has demonstrated significant advancements compared to the current state-of-the-art clustered federated learning approaches.

Microgrids face challenges in monitoring and controlling the power quality (PQ) of integrated electrical systems to make timely decisions. Inverter-based technologies handle small-scale smart grids’ power quality parameters (PQPs) and play an important role in condition monitoring. Accurate forecasting of such parameters is difficult due to the stochastic nature of demand, distributed generation, and weather conditions. Moreover, energy clients have concerns over growing privacy and security breaches for collaboration involving data exchanges. This study aims to predict PQPs indices of home microgrids using ANN, LSTM, and CNN-LSTM models. To preserve users’ privacy, federated learning has been applied with some adaptive differential privacy on the global model and clients’ data. Comparative analysis of the ML model and DP parameters shows that the LSTM model gives better results with adequate privacy parameters to predict the PQPs of five distributed microgrids. LSTM model gives the least MAE of 0.2323 for FL without privacy and 0.3256 test loss for appropriate DP level.

This study aims to predict humor in a binary label, i.e., the presence or absence of humor in video recordings. The challenge here is to predict the variable in a cross-cultural manner, where the training data is in German, and the testing is done on the recordings of English-language-speaking football coaches. The novelty of this paper lies in exploring audio and textual features to predict humor in a cross-cultural setting. It is interesting to study audio and text-based features due to the cross-cultural nature of the problem, which remains largely unexplored when studying pose and facial features. The paper explores several audio (mms-lid, wav2vec 2.0) and textual (LaBSE, multilingual-e5-base) features and then uses them to train both RNN and Transformer encoder architectures. Experiments have been performed on the transformer encoder architectures with and without position encoding to study the effects of the absence of positional encoding in those features for humor detection. Late fusion has also been studied with combinations of all three modalities. We achieved our best AUC Score of 0.9251 and 0.8245 for the development and test set, respectively, out of five given submissions.

The Heart Rate Monitoring system is developed using Artificial Intelligence (ANN) and Internet of Things technology to detect the heartbeat and SpO2 of the patient to monitor the risk of heart attack and also make the person do regular checkups. Health monitoring is very important to us to make sure our health is in which condition. The proposed framework examines the sources of information gathered from sensors fit with the patients to discover the crisis circumstance. The IoT technology is employed; it helps anyone monitor the patient's health from anywhere. A deep learning algorithm such as ANN is utilized to identify whether the person’s health is normal or abnormal. In case an abnormality in health is noticed, an alert will be popped out. The proposed framework with artificial intelligence and IoT can reduce the death occurring due to heart rate, and other related issues can also be avoided.

Digital health care is emerging along with the growth and acceptability of modern ICT technologies such as IoT and Artificial Intelligence. Today’s many health services are being served by computer-related technologies like telemedicine, teleconsultation, and remote diagnosis. However, mental health still lacks ICT technology integration due to privacy and highly sensitive data sharing. Recently, privacy-preserving technologies have been researched and applied to various privacy-required domains. To harness the benefits of these technologies and address the bottleneck of existing solutions in mental health-related solutions, we have proposed a federated learning-based solution for classifying human emotion into seven classes adoration, amusement, anxiety, disgust, empathic pain, and fear and surprise. Our proposed solution preserves data privacy while providing classification accuracy nearly equal to the traditional centralized machine learning solutions.

The amalgamation of the Internet of Things (IoT) and federated learning (FL) is leading the next generation of data usage due to the possibility of deep learning with data privacy preservation. The FL architecture currently assumes labeled data samples from a client for supervised classification, which is unrealistic. Most research works in the literature focus on local training, update receiving, and global model updates. However, by principle, the labeling must be performed on the client side because the data samples cannot leave the source under the FL principle. In the literature, a few works have proposed methods for unlabeled data for FL using “class-prior probabilities” or “pseudo-labeling”. However, these methods make either unrealistic or uncommon assumptions, such as knowing class-prior probabilities are impractical or unavailable for each classification task and even more challenging in the IoT ecosystem. Considering these limitations, we explored the possibility of performing federated learning with unlabeled data by providing a clustering-based method of labeling the sample before training or federation. The proposed work will be suitable for every type of classification task. We performed different experiments on the client by varying the labeled data ratio, the number of clusters, and the client participation ratio. We achieved accuracy rates of 87% and 90% by using 0.01 and 0.03 of the truth labels, respectively.

The new paradigm of nature-inspired cybersecurity can establish a robust defense by utilizing well-established nature-inspired computing algorithms to analyze networks and act quickly. The existing research focuses primarily on the efficient selection of features for quick and optimized detection rates using firefly and other nature-inspired optimization techniques. However, selecting the most appropriate features may be specific to the network, and a different set of features may work better than the selected one. Therefore, there is a need for a generalized pre-processing step based on the standard network monitoring parameters for the early detection of suspicious nodes before applying feature-based or any other type of monitoring. This paper proposes a modified version of the firefly optimization algorithm to effectively monitor the network by introducing a novel health function for the early detection of suspicious nodes. We implement event management schemes based on the proposed algorithm and optimize the observation priority list based on a genetic evolution algorithm for real-time events in the network. The obtained simulation results demonstrate the effectiveness of the proposed algorithm under various attack scenarios. In addition, the results indicate that the proposed method reduces approximately 60–80% of the number of suspicious nodes while increasing the turnaround time by only approximately 1–2%. The proposed method also focuses specifically on accurate network health monitoring to protect the network proactively.

Today machine learning and deep learning are being used in all industrial and social reform. The use and acceptability of AI solutions are rapidly growing. So, traditional agriculture practices adopt these modern technologies and move towards precision farming. Pest detection and classification is one of the critical areas of concern for agriculture and farmers. Deep learning-based detection and classification have recently automated the process, making detection significantly faster. However, centralized training is required to upload crop images (infected or not infected) which leads to privacy invasion and may lead to a negative reputation for the crop. It may incur financial losses due to the low pricing of the harvest to the farmer. Therefore, we provide privacy-preserving pest detection and classification using personalized federated learning that generates detection models based on agricultural characteristics while the farmers keep the data. We provide a performance comparison between centralized and federated approaches for five classes of pests. Further, we perform experiments to create a group-based personalized model. Through experiments, we found that the accuracy of the federated approach is lower than the centralized training. We also found that the performance of groups varies in personalized FL, so the accuracy of Group A (0.69) is higher than Group B (0.63). Based on the experimental result, the proposed solution is suitable for agriculture because of the privacy preservation and personalization with distributed and low computing devices.

Building energy planning is a challenging task in the current mounting climate change scenario because the sector accounts for a reasonable percentage of global end-use energy consumption, with a one-fifth share of global carbon emissions. Energy planners rely on physical model-based prediction tools to conserve energy and make decisions towards decreasing energy consumption. For precise forecasting, such a model requires the collection of an enormous number of input variables, which is time-consuming because not all the parameters are easily available. Utilities are reluctant to share retrievable consumer information because of growing concerns regarding data leakage and competitive energy markets. Federated learning (FL) provides an effective solution by providing privacy preserving distributed training to relieve the computational burden and security concerns associated with centralized vanilla learning. Therefore, we aimed to comparatively analyze the effectiveness of several data-driven prediction algorithms for learning patterns from data-efficient buildings to predict the hourly consumption of the building sector in centralized and FL setups. The results provided comparable insights for predicting building energy consumption in a distributed setup and for generalizing to diverse clients. Moreover, such research can benefit energy designers by allowing them to use appropriate algorithms via transfer learning on data of similar features and to learn personalized models in meta-learning approaches.

Diabetes is a chronic ailment characterized by abnormal blood glucose levels. Diabetes is caused by insufficient insulin synthesis or by cells' insensitivity to insulin activity. Glucose is essential to health since it is the primary source of energy for the cells that make up a person's muscles and tissues. On the condition that if a person has diabetes, his or her body either does not create enough insulin or cannot utilize the insulin that is produced. When there isn't enough insulin or cells stop responding to insulin, many dextroses accumulate in the person's vascular framework. As time passes, this could lead to diseases such as kidney disease, vision loss, and coronary disease. Although there is no cure for diabetes, losing weight, eating nutritious foods, being active, and closely monitoring the diabetes level can all assist. In this research, we used Artificial Neural Network to create a Deep Learning (DL) model for predicting Diabetes. Then it was validated using an accuracy of 92%. In addition, with the help of the MIT website, a mobile application was constructed. This project will now assist in predicting the effects of diabetes and deliver personalized warnings. Early detection of pre-diabetes can be extremely beneficial to patients since studies have shown that symptoms of early diabetic difficulties frequently exist at the time of diagnosis.

Diabetes is a chronic ailment characterized by abnormal blood glucose levels. Diabetes is caused by insufficient insulin synthesis or by cells' insensitivity to insulin activity. Glucose is essential to health since it is the primary source of energy for the cells that make up a person's muscles and tissues. On the condition that if a person has diabetes, his or her body either does not create enough insulin or cannot utilize the insulin that is produced. When there isn't enough insulin or cells stop responding to insulin, many dextroses accumulate in the person's vascular framework. As time passes, this could lead to diseases such as kidney disease, vision loss, and coronary disease. Although there is no cure for diabetes, losing weight, eating nutritious foods, being active, and closely monitoring the diabetes level can all assist. In this research, we used Artificial Neural Network to create a Deep Learning (DL) model for predicting Diabetes. Then it was validated using an accuracy of 92%. In addition, with the help of the MIT website, a mobile application was constructed. This project will now assist in predicting the effects of diabetes and deliver personalized warnings. Early detection of pre-diabetes can be extremely beneficial to patients since studies have shown that symptoms of early diabetic difficulties frequently exist at the time of diagnosis.<br>

In the current ever-changing cybersecurity scenario, active cyber defense strategies are imperative. In this work, we present a standard testbed to measure the efficacy and efficiency of customized networks while analyzing various parameters during the active attack. The presented testbed can be used for analyzing the network behavior in presence of various types of attacks and can help in fine-tuning the proposed algorithm under observation. The proposed testbed will allow users to design, implement, and evaluate the active cyber defense mechanisms with good library support of nature-inspired and AI-based techniques. Network loads, number of clusters, types of home networks, and number of nodes in each cluster and network can be customized. While using the presented testbed and incorporating active-defense strategies on existing network architectures, users can also design and propose new network architectures for effective and safe operation. In this paper, we propose a unified and standard testbed for cyber defense strategy simulation and bench-marking, which would allow the users to investigate current approaches and compare them with others, while ultimately aiding in the selection of the best approach for a given network security situation. We have compared the network performance in difference scenarios namely, normal, under attack and under attack in presence of NICS-based adaptive defense mechanism and achieved stable experimental results. The experimental results clearly show that the proposed testbed is able to simulate the network conditions effectively with minimum efforts in network configuration. The simulation results of defense mechanisms verified on the proposed testbed got the improvement on almost 80 percent while increasing the turnaround time to 1–2 percent. The applicability of proposed testbed in modern technologies like Fog Computing and Edge Computing is also discussed in this paper.

The sudden outbreak of the novel coronavirus disease in 2019, known as COVID-19 has impacted the entire globe and has forced governments of various countries to a partial or full lockdown in the fear of the rapid spread of this disease. The major lesson learned from this pandemic is that there is a need to implement a robust system by using non-pharmaceutical interventions for the prevention and control of new contagious viruses. This goal can be achieved using the platform of the Internet of Things (IoT) because of its seamless connectivity and ubiquitous sensing ability. This technology-enabled healthcare sector is helpful to monitor COVID-19 patients properly by adopting an interconnected network. IoT is useful for improving patient satisfaction by reducing the rate of readmission in the hospital. The presented work discusses the applications and technologies of IoT like smart and wearable devices, drones, and robots which are used in healthcare systems to tackle the Coronavirus pandemic This paper focuses on applications of cognitive radio-based IoT for medical applications, which is referred to as "Cognitive Internet of Medical Things" (CIoMT). CIoMT is a disruptive and promising technology for dynamic monitoring, tracking, rapid diagnosis, and control of pandemics and to stop the spread of the virus. This paper explores the role of the CIoMT in the health domain, especially during pandemics, and also discusses the associated challenges and research directions.

Mental health-related illnesses like depression and anxiety have become a major concern for society. Due to social stigma and unawareness, many such patients lack proper doctors' consultancy, diagnosis, and treatments. Many of such problems arise due to the urban living style, social and family disconnect. Nonetheless, various health-related mobile applications offer ways to minimize the risks by providing an option for remote doctor consultations, connecting to family and friends, or sharing thoughts. To investigate the effectiveness and discover the key issues of such apps, we analyze the user's interaction, discussion, and responses to mental health apps through topic modeling approaches. The experimental results show that many users found many applications helpful and complained mostly about technical issues and business practices. We also evaluated topics and keywords from feedbacks to further improve the application interface and functionalities.

Wireless sensor network (WSN) based applications have become regular in our daily lives, and their integration with the Internet of Things (IoT) makes them even more productive and convenient. However, the inclusion of massive devices, which form the core of IoT, raises plenty of security concerns. Many security strategies to handle data privacy have been proposed in previous work. However, very few of these schemes addressed the confidentiality of the sensor node's position. The goal of this study is to ensure source node location privacy by source location privacy preservation technique by randomized ring routing and confounding transmission (SLP-RRRCT). Random routing with confusing transmission benefits the SLP-RRRCT by distracting the adversary backtracking process. After the random routing phase expires, packets are forwarded to the base station (BS) by neighbour grid-based ring routing. During the development of SLP-RRRCT, we concentrated on the privacy of sensor node position, sensor energy usage, network lifetime, and packet routing randomization. Through a simulation experiment and theoretical analysis, we have observed that the proposed SLP-RRRCT provides better safety time, randomization in transmission delay, and large network lifetime than the compared techniques, i.e., baseline, probabilistic, phantom, source location protection protocol based on dynamic routing (SLPDR), and source location privacy protection scheme based on ring-loop routing (SLPRR).

Integrating traditional critical infrastructure "physical" and Information and Communications Technology (ICT) "cyber" enables many services and can make the cyber-physical system (CPS) more efficient and effective. However, ICT exposes the critical infrastructure outside of its physically closed perimeter, making it vulnerable to cyber-attacks. Although defence systems based upon machine learning algorithms are recently adopted to protect against cyberattacks to CPS, there remains a challenge to choose the most suitable ML system for the underlying CPS. Therefore, this paper benchmarks performance of various machine learning algorithms on four different CPS datasets in two dimensions. First, the performance of machine learning algorithms is benchmarked in terms of accuracy, precision, recall, F1-score, and AUC. Second, the computation requirement is benchmarked during training, prediction, and deployment. Our comprehensive experimental results will help to decide the machine model with the best performance for critical infrastructure with various computation and communication limitations. The experimental result found that a linear model is faster, and bulk prediction is more suitable for CPS. The decision tree is one of the appropriate models considering detection performance and model size.

The idea of federated learning is given by Google in 2016, to develop local-to-global machine learning models by using data locally. Federated learning is a machine learning approach where the objective is to train a centralized machine learning model without aggregating training data from a large number of clients (typically mobile phones). In this approach, each client independently trained the local model by using its local data and sending update to a central server, where the updates from clients are combined to evolve a new high-quality global model. This chapter includes brief overview of the decentralized approach federated learning and its applications.

In recent times, the wireless sensor network (WSN) has become an integral part of daily life. WSN forms the necessary foundation for several important applications such as animal monitoring, border surveillance, asset monitoring, etc. These applications help maintain the confidentiality of the location of the occurring event from the attacker. The properties of the sensor nodes such as limited energy source, communication capability, memory, and network deployment at a large scale make it challenging to maintain the location privacy of a source node. To secure the source node location, this paper presents a source location privacy protection scheme that is based on random rings and a limited hop fake packet routing scheme (SLP-RRFPR). In the proposed scheme, an event packet is forwarded away from the base station by the random routing with confounding transmission to change the attacker’s backtracking process. Afterward, it follows the random routing, where the phantom node forwards the fake packet to other randomly selected nodes. In the last phase, a real packet is transmitted at the base station by ring routing. The simulation results of the proposed SLP-RRFPR are compared with phantom, baseline, probabilistic, source location privacy protection scheme based on ring-loop routing (SLPRR), and source location protection protocol based on dynamic routing (SLPDR). The simulation results show that the proposed SLP-RRFPR performs better than the compared protocol for various performance metrics, such as safety time, transmission delay, network lifetime, and randomness in the packet path.