TOP 1%
Introduction
Initially, local physical devices connected to the Internet for real-time data analysis were considered as the IoT network. Physical limitations would continue and be increased by the demands of recent trends, as the technologies around the edge of the IoT expand rapidly and increase their potential, namely: (i) Transferring data over the Internet to specific online services in a way that standardized enabled by connectivity and then interoperability [6-8]; (ii) the need for higher intelligence at the edge of the network, allowing systems to make choices faster while consuming less energy [9, 10]; (iii) developed security devices, mitigating risks from a large number of massive attack surfaces present in the IoT network [11, 12]; and (iv) new energy saving techniques, allowing autonomous and sustainable devices [13, 14].
IoT edge: trends and challenges
- Basis for connectivity and interoperability
- Edge intelligence
- Security
- Energy awareness
Nevertheless, IoT's diverse set of security challenges and vulnerabilities have an inherent impact on all layers of the architecture. The security layer of the platform is also responsible for verifying the integrity and authenticity of the software executed in the IoT device.
Roles of reconfigurable platforms
- Connectivity and interoperability
- Intelligence
- Security
- Energy
- Combination of reconfigurable platforms and IoT Motes
In addition, the benefits of using FPGA-based cryptographic accelerators have been widely discussed in the literature; Piedra et al. PUF-based applications in [49] serve as a means of security software on an MCU and as a basis for authentication of IoT devices in the cloud.
![Figure 2 illustrates the various software and hardware architectures that are now available on the market and commonly employed in the creation of embedded systems [21]](https://thumb-ap.123doks.com/thumbv2/1libvncom/9200604.0/27.918.136.779.932.1261/illustrates-software-hardware-architectures-available-commonly-employed-creation.webp)
Connected the unconnected world and things: an evolution in connectivity beyond the 5G revolution
On numerous fronts, FPGA SoC-based edge computing attempts to alleviate some of the limitations of cloud computing. All those different factors indicate the upgrading of edge computing around the world.
Proposed QoS-QoR aware CNN FPGA accelerator Co-design approach for feature IoT world
- QoS-QoR CNN accelerator for IoT devices
- Proposed architecture: Acceleration and designing tools
While cloud computing took 10 to 15 years to mature, edge computing is moving faster. Edge computing can be seen as an extension of this move to a more decentralized model.
Results and discussion
Conclusion
Deep Learning and Reconfigurable Platforms in the Internet of Things: Challenges and Opportunities in Algorithms and Hardware. Securing the Internet of Things in the age of machine learning and software-defined networking.
Methodology
- Search approach
- Extracted information
Due to the sensitive nature of the information, the technologies to be implemented are those with features that enable compliance with data privacy and security policies and standards [6]. The rest of the article is divided as follows: part two presents the methodology of the collection and analysis performed, detailing the aspects and characteristics we focus on, part three presents the results obtained, and finally the final section presents the conclusions. For the first purpose of filtering, basic information about these works was collected, in particular the title of the article, abstract, authors, access link to the publication, the number of citations and character of the document according to Google Scholar.
All information analyzes and the filtering process were carried out jointly by the authors and cross-checked. In addition, the institution country supporting each investigation was selected to identify those countries that have the greatest impact worldwide in the IoT for eHealth and telehealth fields.
Results and discussion
- Results obtained after filtering process
- Countries with contribution with higher impact
To a lesser extent, we also find the use of IoT to facilitate the remote diagnosis of the patient. The first of these two areas mentioned was the one that had the most influence in the beginning of the use of IoT in healthcare, being currently on the slope of enlightenment or productivity plateau in the hype cycle curve. In the results, we also observe that the studies related to the study of interoperability and the analysis of user friendliness, user experience and the degree of acceptance have very little effect.
The lack of research on these issues may be the biggest limitation of these systems in the future. This result again reveals indications of the situation of these systems in the hype cycle curve, as reduced interest to delve into new applications and consolidate their use by focusing on the largest.
Conclusions
Cloud and IoT based disease prediction and diagnosis system for health care using fuzzy neural classifier. Privacy-preserving smart IoT-based healthcare big data storage and self-adaptive access control system. Towards a remote monitoring of patient vital signs based on IoT-based blockchain integrity management platforms in smart hospitals.
COVID-SAFE: An Internet of Things based system for automated health monitoring and surveillance in post-pandemic life. A secure authentication and encryption framework using enhanced ECC for IoT-based medical sensor data.
Related work
- Routing protocols in IoT
- Intrusion detection systems in IoT
Still, the advanced ID schemes that use machine learning techniques struggle to detect some cyber-attacks. Here we study the potential applicability of the hierarchical hidden Markov model (HHMM) for intrusion detection in IoT systems where the problem space can be several orders larger than in wireless networks. The proposed scheme shows better results in detecting the DoS and DDoS attack patterns compared to the advanced work.
Due to the lack of training data sets, current IoT intrusion detection systems are unable to detect the latest DoS and DDoS attacks [10], such as Network Time. They proposed a multi-stage Naive Bayes model that can predict each stage of the multi-stage attack scenarios.
Preliminaries
- Distributed denial of service attacks (DDoS)
- Hierarchical hidden Markov model (HHMM)
Moreover, the authors in [13] propose a Hierarchical Hidden Markov Model (HHMM), which is an extension of the hidden Markov model (HMM), as the method for activity recognition. Their proposed model identifies the attacks by the back-propagation neural network model. Table 1 summarizes the characteristics of the main types of existing ID schemes. DDoS can still cause a long-term memory consumption of the redeployment nodes in IoT environments due to nodes' limited resources.
Many policy algorithms use different measurements such as standard deviation or measure the chi-square statistic of the sample to classify the packets as malicious or legitimate. The Hierarchical Hidden Markov Model (HHMM) is a multilevel stochastic process derived from the Hidden Markov Model (HMM) by making each of the hidden states a self-contained autonomous probability model.
Framework of the proposed model
- Framework of the HHMM
- Framework of PHHMM
This probabilistic hierarchical hidden Markov model should overcome the problem of IoT data heterogeneity. Estimate the model parameters: find the most likely parameterλ∗ of the model by applying the Baum-Welch algorithm [19]: given the PHHMM structure and one or more observation seriesλ¼argmaxλ∗P Oð tjλÞ. Evaluation: Detect DDoS attacks: Evaluate the probability of the observed sequence and solve the detection problem to calculate the probability of alert observations by applying the DDoS detection algorithm: given the probability of an observed attack alert sequence (O), detect DDoS attacks and predict future paths.
The participating nodes in the algorithm use the PCA with the SVD learning mechanism to estimate the principal components of the data traffic. By using only the most important principal components, we could avoid the calculation of the entire subspace.
Proposed model
- Data pre-processing
- Feature selection
- Data clustering
- Dimensionality reduction
In our PHHMM model, the application of dimensionality reduction techniques is a challenging step due to the lack of a standard approach for dimensionality reduction of observed IoT network traffic. Determining the principal components with respect to the covariance matrix is computationally expensive as it requires an eigenvalue decomposition that requires the computation of the covariance matrix. CICIDS2019 includes inbound and outbound traffic of the most advanced DoS and DDoS attacks.
This order stores the elements in order of weight relative to the variance of the initial data matrix. The eigenvectors are called the principal axes of the data, and the data projections onto the principal axes are called principal components [24].
Dimensionality Reduction Algorithm
- Hierarchical hidden Markov classification
- Initialize Observations (M), States (q i ), Threshold (Th)
Similar to HHMM [17], the PHHMM model uses the Baum-Welch algorithm to calculate the likelihood-maximizing parameters of the model given the observed data. Then, the model uses the Viterbi algorithm to find the most probable sequence of the hidden states given the observed data and the parameters. Learning: The first-level observations train L1LHMMi using the Baum-Welch algorithm to determine model parameters.
Decoding: The observations of the second level train the L2LHMMiby Viterbi algorithm to find the most probable sequence of hidden states using model. Evaluation: The observations of the third level train the L3LHMMi and find the DDoS attack sequence using the most probable sequence.
Viterbi algorithm [26]
This algorithm uses only state-optimized joint likelihood for observation data and the underlying Markovian state sequence as the objective function for estimation. Unlike the BW algorithm, it does not update all probable paths for all states in the HHMM. Pattern Detection Algorithm This algorithm uses prior knowledge to learn about the past attack behavior and track the attack alerts.
Obtain the probability of the observation sequence possibilityOqid then predicts the DDoS attack behavior based on the appearance of the attack observation sequence in the previous algorithm.
Detection algorithm
- Experimental setup 1 Datasets
- Evaluation metrics
- Evaluation results
- Classification accuracy
- Efficiency
- Conclusion
- Related research
- Datasets with classification labels
- Types and placement of sensors
- Research methodology
- Implementation procedure, software, and hardware
- Sensors
- Datasets
- Algorithms selection from common algorithms
- Design of the floor plan of the smart house
- Remainder alarm for medicine taking routine and camera/pressure mat There would be a remainder device. This becomes the third Component of the
- Description of the three algorithms
Nine of the common classical ML algorithms are k-Means, Linear Discriminant Analysis (LDA), Naïve-Bayes, K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Artificial Neural Network (ANN), Random Forest and Decision trees [8]. Environmental and wearable sensors are combined to identify the location and the movement of the subject person. The performances of the algorithms were compared with the dataset sub-class of these two datasets.
Below is Table 2 showing the composition of the data set that will be used in the experiment. We would compare these algorithms and use the best one in the smart house application.
Update datasets, test, and use them in future training
5. Staying on the toilet for a long time indicates a problem, therefore the alarm is evoked, 6. A medical dispensary is kept in the room (E). We must perform two tests before sending an alert as a way to avoid false alarms. Identify the activity and then identify the room and location of the applicable algorithm also determine that the layout is not in an inappropriate room.
Sleeping area locator for logical heuristic missed fall prediction
Medical remainder algorithm, to predict skipping of medicine routine
- Updating training dataset
- Results
- Discussions and future works
- Edge processing and security
- Replacement of training and testing dataset
- The medicine routine remainder service
- Conclusion
- Materials and methods
- Taxonomy
- Intelligent instrument scenarios
- Audio processing scenarios
- Music generator scenarios
- Music recommendation device scenarios .1 Emotiwatch scenario
- Feedback device scenarios .1 RumbleRumble scenario
- Educational scenarios
- Processing architecture
- Results
- Scenario implementation
- TherAImin
- Discussion
- Conclusions
Below is Table 4 showing the performance of the most effective classifier (XGboost) in the experiment. The overall accuracy for the xgboost algorithm was above 96% on each of the three datasets. The sample data should also be replaced with the change in the pattern of the senior citizen trends.
In this way, the device can also be considered as part of the Internet of Behavior (IOB) [7]. In this section, we present a possible implementation for some of the devices proposed in the scenarios. The rest of this layer consists of the video interface already available on the raspberry pi.
TherAImin, as an extension of the Theremin, can be considered in the class of musical instruments added to the FAIME taxonomy.
