RESUMO
Machine learning (ML) represents one of the main pillars of the current digital era, specifically in modern real-world applications. The Internet of Things (IoT) technology is foundational in developing advanced intelligent systems. The convergence of ML and IoT drives significant advancements across various domains, such as making IoT-based security systems smarter and more efficient. However, ML-based IoT systems are vulnerable to lurking attacks during the training and testing phases. An adversarial attack aims to corrupt the ML model's functionality by introducing perturbed inputs. Consequently, it can pose significant risks leading to devices' malfunction, services' interruption, and personal data misuse. This article examines the severity of adversarial attacks and accentuates the importance of designing secure and robust ML models in the IoT context. A comprehensive classification of adversarial machine learning (AML) is provided. Moreover, a systematic literature review of the latest research trends (from 2020 to 2024) of the intersection of AML and IoT-based security systems is presented. The results revealed the availability of various AML attack techniques, where the Fast Gradient Signed Method (FGSM) is the most employed. Several studies recommend the adversarial training technique to defend against such attacks. Finally, potential open issues and main research directions are highlighted for future consideration and enhancement.
RESUMO
This paper proposes a high-accuracy EEG-based schizophrenia (SZ) detection approach. Unlike comparable literature studies employing conventional machine learning algorithms, our method autonomously extracts the necessary features for network training from EEG recordings. The proposed model is a ten-layered CNN that contains a max pooling layer, a Global Average Pooling layer, four convolution layers, two dropout layers for overfitting prevention, and two fully connected layers. The efficiency of the suggested method was assessed using the ten-fold-cross validation technique and the EEG records of 14 healthy subjects and 14 SZ patients. The obtained mean accuracy score was 99.18 %. To confirm the high mean accuracy attained, we tested the model on unseen data with a near-perfect accuracy score (almost 100 %). In addition, the results we obtained outperform numerous other comparable works.
RESUMO
Internet of Drones (IoD) plays a crucial role in the future Internet of Things due to its important features such as low cost, high flexibility, and mobility. The number of IoD applications is drastically increasing from military to civilian fields. Nevertheless, drones are resource-constrained and highly vulnerable to several security threats and attacks. The use of blockchain technology for securing IoD networks has gained growing attention. To this end, this paper presents a systematic literature review to analyze the current research area regarding the security of IoD environments using the emerging blockchain technology. Forty relevant studies were selected from 129 published articles to answer the identified research questions. The selected studies were classified into three main classes based on blockchain type. Furthermore, a comparison of the reviewed articles in terms of different factors is provided. The research findings show that the blockchain can guarantee fundamental security requirements such as authentication, privacy-preserving, confidentiality, integrity, and access control. Finally, open issues and challenges related to the combination of blockchain and IoD technologies are discussed.
RESUMO
Limitations of available literature: Nowadays, coronavirus disease 2019 (COVID-19) is the world-wide pandemic due to its mutation over time. Several works done for covid-19 detection using different techniques however, the use of small datasets and the lack of validation tests still limit their works. Also, they depend only on the increasing the accuracy and the precision of the model without giving attention to their complexity which is one of the main conditions in the healthcare application. Moreover, the majority of healthcare applications with cloud computing use centralization transmission process of various and vast volumes of information what make the privacy and security of personal patient's data easy for hacking. Furthermore, the traditional architecture of the cloud showed many weaknesses such as the latency and the low persistent performance. Method proposed by the author with technical information: In our system, we used Discrete Wavelet transform (DWT) and Principal Component Analysis (PCA) and different energy tracking methods such as Teager Kaiser Energy Operator (TKEO), Shannon Wavelet Entropy Energy (SWEE), Log Energy Entropy (LEE) for preprocessing the dataset. For the first step, DWT used to decompose the image into coefficients where each coefficient is vector of features. Then, we apply PCA for reduction the dimension by choosing the most essential features in features map. Moreover, we used TKEO, SHEE, LEE to track the energy in the features in order to select the best and the most optimal features to reduce the complexity of the model. Also, we used CNN model that contains convolution and pooling layers due to its efficacity in image processing. Furthermore, we depend on deep neurons using small kernel windows which provide better features learning and minimize the model's complexity.The used DWT-PCA technique with TKEO filtering technique showed great results in terms of noise measure where the Peak Signal-to-Noise Ratio (PSNR) was 3.14 dB and the Signal-to-Noise Ratio (SNR) of original and preprocessed image was 1.48, 1.47 respectively which guaranteed the performance of the filtering techniques.The experimental results of the CNN model ensure the high performance of the proposed system in classifying the covid-19, pneumonia and normal cases with 97% of accuracy, 100% of precession, 97% of recall, 99% of F1-score, and 98% of AUC. Advantages and application of proposed method: The use of DWT-PCA and TKEO optimize the selection of the optimal features and reduce the complexity of the model.The proposed system achieves good results in identifying covid-19, pneumonia and normal cases.The implementation of fog computing as an intermediate layer to solve the latency problem and computational cost which improve the Quality of Service (QoS) of the cloud.Fog computing ensure the privacy and security of the patients' data.With further refinement and validation, the IFC-Covid system will be real-time and effective application for covid-19 detection, which is user friendly and costless.
RESUMO
Under a dense and large IoT network, a star topology where each device is directly connected to the Internet gateway may cause serious waste of energy and congestion issues. Grouping network devices into clusters provides a suitable architecture to reduce the energy consumption and allows an effective management of communication channels. Although several clustering approaches were proposed in the literature, most of them use the single-hop intra-clustering model. In a large network, the number of clusters increases and the energy draining remains almost the same as in un-clustered architecture. To solve the problem, several approaches use the k-hop intra-clustering to generate a reduced number of large clusters. However, k-hop proposed schemes are, generally, centralized and only assume the node direct neighbors information which lack of robustness. In this regard, the present work proposes a distributed approach for the k-hop intra-clustering called Distributed Clustering based 2-Hop Connectivity (DC2HC). The algorithm uses the two-hop neighbors connectivity to elect the appropriate set of cluster heads and strengthen the clusters connectivity. The objective is to optimize the set of representative cluster heads to minimize the number of long range communication channels and expand the network lifetime. The paper provides the convergence proof of the proposed solution. Simulation results show that our proposed protocol outperforms similar approaches available in the literature by reducing the number of generated cluster heads and achieving longer network lifetime.
RESUMO
BACKGROUND AND OBJECTIVES: Recent advances in miniature biomedical sensors, mobile smartphones, wireless communications, and distributed computing technologies provide promising techniques for developing mobile health systems. Such systems are capable of monitoring epileptic seizures reliably, which are classified as chronic diseases. Three challenging issues raised in this context with regard to the transformation, compression, storage, and visualization of big data, which results from a continuous recording of epileptic seizures using mobile devices. METHODS: In this paper, we address the above challenges by developing three new algorithms to process and analyze big electroencephalography data in a rigorous and efficient manner. The first algorithm is responsible for transforming the standard European Data Format (EDF) into the standard JavaScript Object Notation (JSON) and compressing the transformed JSON data to decrease the size and time through the transfer process and to increase the network transfer rate. The second algorithm focuses on collecting and storing the compressed files generated by the transformation and compression algorithm. The collection process is performed with respect to the on-the-fly technique after decompressing files. The third algorithm provides relevant real-time interaction with signal data by prospective users. It particularly features the following capabilities: visualization of single or multiple signal channels on a smartphone device and query data segments. RESULTS: We tested and evaluated the effectiveness of our approach through a software architecture model implementing a mobile health system to monitor epileptic seizures. The experimental findings from 45 experiments are promising and efficiently satisfy the approach's objectives in a price of linearity. Moreover, the size of compressed JSON files and transfer times are reduced by 10% and 20%, respectively, while the average total time is remarkably reduced by 67% through all performed experiments. CONCLUSIONS: Our approach successfully develops efficient algorithms in terms of processing time, memory usage, and energy consumption while maintaining a high scalability of the proposed solution. Our approach efficiently supports data partitioning and parallelism relying on the MapReduce platform, which can help in monitoring and automatic detection of epileptic seizures.