|Computer Systems Science & Engineering |
Neural Cryptography with Fog Computing Network for Health Monitoring Using IoMT
1Department of Computer Science and Engineering, Coimbatore Institute of Engineering and Technology, Coimbatore, 641109, India
2Department of Applied Cybernetics, Faculty of Science, University of Hradec Králové, 50003, Hradec Králové, Czech Republic
3Department of Information Technology, College of Computers and Information Technology, Taif University, P.O. Box 11099, Taif, 21944, Saudi Arabia
4Department of Computer Science, College of Computers and Information Technology, Taif University, P. O. Box 11099, Taif, 21944, Saudi Arabia
5Department of Mathematics, Faculty of Science, Mansoura University, Mansoura, 35516, Egypt
6Department of Computational Mathematics, Science, and Engineering (CMSE), Michigan State University, East Lansing, MI, 48824, USA
*Corresponding Author: Mohamed Abouhawwash. Email: email@example.com
Received: 24 October 2021; Accepted: 24 January 2022
Abstract: Sleep apnea syndrome (SAS) is a breathing disorder while a person is asleep. The traditional method for examining SAS is Polysomnography (PSG). The standard procedure of PSG requires complete overnight observation in a laboratory. PSG typically provides accurate results, but it is expensive and time consuming. However, for people with Sleep apnea (SA), available beds and laboratories are limited. Resultantly, it may produce inaccurate diagnosis. Thus, this paper proposes the Internet of Medical Things (IoMT) framework with a machine learning concept of fully connected neural network (FCNN) with k-nearest neighbor (k-NN) classifier. This paper describes smart monitoring of a patient’s sleeping habit and diagnosis of SA using FCNN-KNN+ average square error (ASE). For diagnosing SA, the Oxygen saturation (SpO2) sensor device is popularly used for monitoring the heart rate and blood oxygen level. This diagnosis information is securely stored in the IoMT fog computing network. Doctors can carefully monitor the SA patient remotely on the basis of sensor values, which are efficiently stored in the fog computing network. The proposed technique takes less than 0.2 s with an accuracy of 95%, which is higher than existing models.
Keywords: Sleep apnea; polysomnography; IOMT; fog node; security; neural network; KNN; signature encryption; sensor
The Internet of Medical Things (IoMT) is the amalgamation of medical applications using sensor devices to connect with health-related information. In this scenario, user health data are sensed by the IoMT and transferred to the chief physician through a modern communication system. Even without the physical presence of the patient, the doctor can naturally view his or her health condition to prescribe medication. Thus, transferring and storing the sensitive data must be secured.
Sleep is one of the fundamental daily needs and is significant for brain function. Any sleep disorder can naturally affect general health and cause serious health problems, such as brain stroke, high blood pressure, complicate daily activities, and risk safety. Experts use Polysomnography (PSG) to study sleep disorders. PSG is a collection of signals recorded from various sensors counting the Electroencephalogram (EEG), Electro oculography (EOG), Electromyogram (EMG), Electrocardiogram (ECG), airflow, thoracic and abdominal movements, and oximetry. Rechtschaffen and Kales standardized the sleep stage classification rules depending on EEG modifications and split Non-Rapid Eye Movement (NREM) in sleep is a natural relief of humans and prevents several diseases. Sleep apnea is a sleep disorder in which breathing problem occurs during sleep states. SA affects our health by decreasing the oxygen blood level, eventually leading to other diseases . SA impacts our health, mentally and physically, causing diseases, such as stroke, cardio problems, and diabetes. Scientifically, based upon the international sleep expert’s key observation, roughly one billion people worldwide have obstructive sleep apnea (OSA) .
IoMT collects significant health-related data and helps early diagnosis of diseases using machine learning techniques. Persons health data is considered extremely sensitive and confidential to users. Thus, IoMT must ensure user privacy  through a federated learning technique. This learning method trains the device to stop sharing the data outside the device . For SA diagnosis, PSG test is needed, which produces electrical energy signals from sensor devices attached to the human body. The process of sleep monitoring people with SA is highly expensive and beds remain limited. Sometimes, a full night or two nights are required for monitoring in a well-equipped sleep laboratory. PSG-recorded values sometimes misguide health professionals, leading to inaccurate diagnosis.
However, monitoring a person’s health condition continuously is extremely difficult. Nonetheless, IoMT makes this possible. Although data transfer is performed remotely, its security and privacy are not assured. Moreover, data are sensitive, and a need for authentication technology persists, given that various attacks can enter the IoMT environment through the internet. Some attackers easily acquire control over the medical device remotely. Malware in IoMT greatly affects its communication and control on the medical device. Moreover, traditional techniques are insufficient to highly secure IoMT communication. Recent attacks by Mirai botnets create distributed denial of service due to insecurity in the Internet of Things (IoT) environment. Hence, a strong security technique to detect attacks in the sensitive IoMT environment is warranted [5–7].
This research contributes security features as follows:
1. Sleep apnea is monitored using health sensors, such as an oximeter, and stress is monitored on the basis of the heart rate. Data are collected and transmitted to fog nodes using the IoMT framework.
2. IoMT uses forward neural network to train the features and uses KNN to classify SA accurately. The classified data is stored in fog nodes using a signature encryption technique.
The rest of the article is structured as follows. Section 2 presents the survey on IoMT. Section 3 demonstrates the proposed architecture and apnea detection using IoMT. Section 4 discusses the result evaluation. Section 5 concludes.
2 Literature Survey
The information or data is connected and communicated through the internet using the IoT. Smart devices of homes, cars, watches, and so on are connected to the internet to transfer real-time data from user to end user. Moreover, IoT produces Wireless Sensor Networks (WSN), smart technologies, cloud network, and so on . Human to machine communication  is abruptly transferred to machine-machine communication. Recently the IoT became promising in smart applications, such as the health industry, smart city, and so on. Thus, this survey focused on intelligent monitoring of SA diagnosis continuously. The Internet of Intelligent Things (IoIT) [10–12] blends the artificial intelligence in the healthcare system to analyze the data logically.
The main disadvantages of using the IoT is limited bandwidth processing, less memory, and small structure, which cause security attacks and privacy threats [13–16]. These factors influence high research interest in IoT security. To address IoT security, various cryptographic frameworks and technology are suggested [17–22]. Especially in the medical system, handling patient data is highly sensitive for the IoT [23,24]. Central cloud server requires high security structure for providing data authentication.
Some studies explored the adaptive architecture for IoMT architecture . This work provides high security for health dataset using public and private key structures. However, the security of the system is not highly adaptive. Thus, this work explores whether the hybrid security system using a cyber security algorithm for the IoMT is implemented . Streaming data is transmitted via nodes using high authentication  to secure the IoMT real-time data in networks. Previous research discussed the security of IoMT with different techniques .
This study discussed several SA surveys. The novel recurrent neural architecture is used to extract the apnea features from input dataset . Machine learning classifiers, such as the support vector machine (SVM) , threshold-based detectors , and regression trees with Adaboost are used to classify SA from the input feature set. The health-based smart contracts [32,33] are used for IoMT to transfer health data [34–38].
3 Proposed Methodology
Although SA syndrome is a common disease, patient diagnosis is difficult. Thus, smart monitoring of the sleeping habit of patients is presented. It monitors the variations in stress and observes the patient’s heart rate and the blood oxygen level while asleep. The architecture of SA syndrome analysis is presented in Fig. 1. The sensor transmits the values to the data analysis section. Here, SA is analyzed on the basis of the input SpO2 value. The forward neural network and KNN classifier is used to detect SA in patients.
3.1 Visualizing of Sleep Disorder
During sleep state, the nervous system becomes inactive, and the relaxation of muscles forces the eyes to close. It is also associated with low movement, stress less posture, and consciousness becomes suspended involuntarily. To monitor the sleep disorder, patients must undergo five stages, which include rapid eye movement (REM) and non rapid eye movement (NREM). Fig. 2 shows the detailed sleeping stages categorically.
Human beings have five stages of sleep. Stages 1 and 2 are considered light sleep in which people can be easily awakened. Stage 3 is the deep sleep, Stage 4 represents very deep sleep, and Stage 5 is the REM. Each stage requires 5–15 minutes to complete. After completion of five stages, it starts again from Stage 1. On average, 90 to 110 minutes are needed to complete a sleep cycle. Any irregularity that occurs within the five sleep stages is a sleep disorder. Thus, to maintain a good sound sleep, sufficient sleep in each stage and sleep cycle must be obtained. Fig. 2 presents the sleep stages.
3.2 Monitoring Sleep Apnea
During the sleep state, some breathing-related disorder occurred is called SA. As shown in Fig. 3 for monitoring SA, SpO2 and heart rate variability(HRV) sensor devices are attached in various organs in the body. Electrical energy emitted by the body is collected from these sensors and transmitted as graphical representation and stored in the fog computing network. This procedure is the Polysomnogram diagnosis.
3.3 Proposed - IoMT-Fog Computing
IoMT sensor values, such as the heart rate, and oximetric values are collected and stored in the fog computing-based network. Fig. 4 shows the IoMT with the fog computing process. Fig. 4 contains two layers, namely, the IoMT section and the fog computing layer. In the IoMT layer, data are collected from sensor devices and are classified into sensitive data (confidential data) and non-sensitive data (non-confidential data). This classified data is transferred to the fog layer, which contains fog servers. Then, data encryption and authorization are performed by a hashing technique of Advanced Signature-Based Encryption (ASE) of Diffie-Hellman key exchange and digital signature. This algorithm is used to exchange information between IoMT devices securely.
3.3.1 IoMT Layer
The IoMT layer contains wearable sensor devices of SpO2, which extract the electrical energy signal of SpO2 and HRV. This combination of SpO2 levels and HRV is used to reduce the false detective cases and increases accuracy. This work implementation is focused on monitoring heart rate and oxygen blood level when SA transpires in the sleep state of a patient. When SA happens to the patient, the system will alarm the doctor and observe the readings stored in the fog node securely. The Apnea-Hypopnea Index (AHI) is used to measure SA severity. Tab. 1 displays the AHI value.
Algorithm 1 represents the patient’s SA severity.
In the IoMT layer, data are collected and classified on the basis of the AHI value. The data are classified as sensitive information and non-sensitive information using FCNN and k-nearest neighbour (k-NN) classifier. Fig. 5 shows the FCNN diagram for SA classification.
This FCNN has one input layer, two hidden layers, and one output layer with 20 neurons. This model is used to develop a relationship between AHI value and sleep stages. To obtain accurate classification after applying FCNN, KNN is implemented. In the KNN classification, the input value is the k-nearest neighbor, and the output value assigns the feature vector class membership. In the classification of KNN, Euclidean distance metric measures are used,
In Eq. (2), classification is done by the majority votes of neighbourhood values that are grouped together. Here the value of k should be assigned only an odd number. That is, 1-NN, 3-NN, 5NN, and so on.
3.3.2 Fog Layer
In the fog layer, the classification of data is stored for promoting accuracy, security, and scalability. Fog servers are used to generate data encryption and decryption using Diffie-Hellman encryption in the fog node. Algorithms 2 and 3 are used to encrypt and decrypt the information.
Algorithm 1 retrieves data from IoMT devices and transfers to fog nodes. In the fog node data from sensor devices of SoP2, HRV are stored securely. Algorithm 2 describes the data encryption and decryption using the Diffie-Hellman encryption during fog computation. ASE algorithm comprises asymmetric cryptographic operations, such as the Diffie-Hellman key exchange and digital signature.
Algorithms 3 and 4 typically perform the encryption and decryption of the SA patient’s data by using the digital signature of hash, and private key produces the Diffie-Hellman method for the exchange of the key among various mobile IoMT devices and fog nodes.
4 Results and Analysis
Independent analysis of SA is a breathing disorder that occurs during a person’s sleep state. In this work, FCNN-KNN+ASE uses two public datasets of Sleep-EDF-2013 and Sleep-EDF-2018. For evaluation, a 20-fold cross validation was applied to all 250 PSG data values of the dataset. The sleeping stages of R & K standard of W, N1, N2, N3, N4, and REM are used to evaluate FCNN-KNN+ASE. The parametric measures are given below.
where TP-True Positive, TN-True Negative, FP-False Positive, FN-False Negative, and is Hypothetical probability. Tabs. 2 and 3 shows that confusion matrix of Sleep-EDF-2013 dataset in FCNN-KNN+ASE using 20-fold cross-validation.
Tab. 4 shows the comparison of our proposed work with other machine learning algorithms in the parametric metric measures of kappa and MF1 score values using two different datasets of Sleep-EDF-2013 and Sleep-EDF-2018.
Tab. 5 unveils that the evaluation of FCNN-KNN+ASE was compared with FCNN-ASE and KNN-ASE. Our proposed work obtained better value in kappa of 0.7 in the Sleep-EDF-2013 dataset and obtained a kappa value of 0.8 in the Sleep-EDF-2018 dataset. Similarly, for the MF1 score value of the Sleep-EDF-2013 dataset obtained 81.37% and the Sleep-EDF-2018 dataset obtained 84.65% in our proposed work, FCNN-KNN+ASE. Tabs. 6 and 7 show the calculating error rate of the Sleep-EDF-2013 and Sleep-EDF-2018 datasets.
Tabs. 6 and 7 present tthat the error rate for our proposed work produces minimum error rate using two different datasets. Tabs. 8 and 9 unveil that cost of communication and storage cost in bits of different algorithm with two different datasets are given.
Tab. 8 demonstrates that our proposed technique obtains minimum communication cost and storage cost compared with other techniques.
Tab. 9 presents that our proposed technique obtains minimum communication cost and storage cost compared with other techniques. Tabs. 9 and 10 reveal that time analysis for encryption of PSG information with various file sizes and stored it in fog-node.
Tab. 10 shows that our proposed work obtained minimum time analysis for encryption using the Sleep-EDF-2013 dataset and the PSG information, which are collected from SoP2 and HRV sensor devices for diagnosis of SAS.
Tab. 10 demonstrates that our proposed work obtained minimum time analysis for encryption using the Sleep-EDF-2018 dataset and the PSG information, which are collected from SoP2 and HRV sensor devices for SAS diagnosis.
Our proposed work obtained minimum time analysis for decryption using the Sleep-EDF-2018 dataset and the PSG information, which are collected from SoP2 and HRV sensor devices for SAS diagnosis. Fig. 6 shows the average accuracy rate for FCNN-KNN+ASE in the two datasets.
Fig. 6 presents that our proposed work gives high accuracy rate compared with other algorithms. Fig. 7 shows the computation time of our proposed work.
The figure shows that the computation of our proposed work obtained minimum time. It produces minimum execution time and minimum error rate.
This work provides reference for SA diagnosis using a fog computing-based IoMT in machine learning algorithms. It uses electrical energy signals that are carefully collected from SpO2 and HRV sensor devices and collected data stored in fog computing network. Thereafter, we implement a complex ASE algorithm for encryption and decryption for PSG values, which cannot be accessed by unauthorized users. This work uses two different public datasets of Sleep-EDF-2013and Sleep-EDF-2018. The advantages of the proposed work, FCNN-KNN +ASE, are better security, faster access, minimal time execution, higher accuracy, more flexible, and more reliable compared with other existing algorithms. The overall accuracy performance of this work obtained 92.13% in the Sleep-EDF-2013 dataset and 93.32% in the Sleep-EDF-2018 dataset. Finally, our proposed work analysis provides effective monitoring for diagnosing people with the SA based on PSG values. Future work is upgraded by using various machine learning algorithms with edge computing. Energy efficiency must be calculated in future work to widely enhance fog computing the IoMT.
Acknowledgement: We would like to give special thanks to Taif University Research supporting Project Number (TURSP-2020/98), Taif University, Taif, Saudi Arabia.
Funding Statement: Taif University Researchers Supporting Project Number (TURSP-2020/98), Taif University, Taif, Saudi Arabia.
Conflicts of Interest: The authors declare that they have no conflicts of interest to report regarding the present study.
|This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.|