Search

Indoor fires may cause casualties and property damage, so it is important to develop a system that predicts fires in advance. There have been studies to predict potential fires using sensor values, and they mostly exploited machine learning models or recurrent neural networks. In this paper, we propose a stack of Transformer encoders for fire prediction using multiple sensors. Our model takes the time-series values collected from the sensors as input, and predicts the potential fire based on the sequential patterns underlying the time-series data. We compared our model with traditional machine learning models and recurrent neural networks on two datasets. For a simple dataset, we found that the machine learning models are better than ours, whereas our model gave better performance for a complex dataset. This implies that our model has a greater potential for real-world applications that probably have complex patterns and scenarios.

Cluster validity indices (CVIs) for evaluating the result of the optimal number of clusters are critical measures in clustering problems. Most CVIs are designed for typical data-type objects called certain data objects. Certain data objects only have a singular value and include no uncertainty, so they are assumed to be information-abundant in the real world. In this study, new CVIs for uncertain data, based on kernel probabilistic distance measures to calculate the distance between two distributions in feature space, are proposed for uncertain clusters with arbitrary shapes, sub-clusters, and noise in objects. By transforming original uncertain data into kernel spaces, the proposed CVI accurately measures the compactness and separability of a cluster for arbitrary cluster shapes and is robust to noise and outliers in a cluster. The proposed CVI was evaluated for diverse types of simulated and real-life uncertain objects, confirming that the proposed validity indexes in feature space outperform the pre-existing ones in the original space.

Data created at industrial sites through industrial internet of things devices are now being processed automatically or in real-time in the industrial structure, due to the application of artificial intelligence technology to industrial sites. However, the expenses of autonomous or real-time data processing and steady data processing (analysis, prediction, prescription, and implementation) necessitate a new processing method. We propose a blockchain-based industrial internet of things information reinforcement model in this work that may reliably ensure the integrity of industrial internet of things data produced at industrial locations. The proposed model processes industrial internet of things data that may occur at endpoints at industrial sites into the blockchain by processing data generated by the same industrial internet of things device independently. As a result, the IIoT data sent to the industrial internet of things server can be evaluated more readily, and production accuracy may be enhanced. The proposed model optimizes industrial internet of things information linkage by stochastically reflecting the information based on attribute value frequency. By dynamically aggregating the related data of industrial internet of things information acquired as a seed through hierarchical subnets, the proposed model increases stability and accuracy. Furthermore, the proposed model may be used to enhance an organizations' operational efficiency (consulting and training, for example) and strategic decision-making by utilizing fundamental knowledge about items produced at industrial locations. Furthermore, the proposed model allows for information sharing and system connectivity between industrial locations, allowing for close collaboration between industrial internet of things features. As a result of the performance evaluation, the proposed model included an industrial internet of things sensor to the blockchain, eliminating the need for an extra function in the manufacturing process and reducing the time required to validate the integrity of industrial internet of things data. In addition, as a result of analyzing industrial internet of things data by an algorithm according to the number of simulated clouds, the accuracy of industrial internet of things information was improved by 2.5% to 3%, on average.

Biological agents used in biological warfare or bioterrorism are also present in bioaerosols. Prompt identification of a biological weapon and its characteristics is necessary. Herein, we optimized an environmentally adaptive detection algorithm that can better reflect changes in the complex South Korean environment than the current models. The algorithm distinguished between normal and biological particles using a laser-induced fluorescence-based biological particle detector capable of real-time measurements and size classification. We ensured that the algorithm operated with minimal false alarms in any environment by training based on experimental data acquired from an area where rainfall, snow, fog and mist, Asian dust, and water waves on the beach occur. To prevent time and money wastage due to false alarms, the detection performance for each level of sensitivity was examined to enable the selection of multiple sensitivities according to the background, and the appropriate level of sensitivity for the climate was determined. The basic sensitivity was set more conservatively than before, with a 3% alarm rate at 20 agent-containing particles per liter of air (ACPLA) and a 100% alarm rate at 63 ACPLA. The reliability was increased by optimizing five variables. False alarms did not occur in situations where no alarm was unnecessary.

As IoT (Internet of Things) devices are diversified in the fields of use (manufacturing, health, medical, energy, home, automobile, transportation, etc.), it is becoming important to analyze and process data sent and received from IoT devices connected to the Internet. Data collected from IoT devices is highly dependent on secure storage in databases located in cloud environments. However, storing directly in a database located in a cloud environment makes it not only difficult to directly control IoT data, but also does not guarantee the integrity of IoT data due to a number of hazards (error and error handling, security attacks, etc.) that can arise from natural disasters and management neglect. In this paper, we propose an optimized hash processing technique that enables hierarchical distributed processing with an n-bit-size blockchain to minimize the loss of data generated from IoT devices deployed in distributed cloud environments. The proposed technique minimizes IoT data integrity errors as well as strengthening the role of intermediate media acting as gateways by interactively authenticating blockchains of n bits into n + 1 and n - 1 layers to normally validate IoT data sent and received from IoT data integrity errors. In particular, the proposed technique ensures the reliability of IoT information by validating hash values of IoT data in the process of storing index information of IoT data distributed in different locations in a blockchain in order to maintain the integrity of the data. Furthermore, the proposed technique ensures the linkage of IoT data by allowing minimal errors in the collected IoT data while simultaneously grouping their linkage information, thus optimizing the load balance after hash processing. In performance evaluation, the proposed technique reduced IoT data processing time by an average of 2.54 times. Blockchain generation time improved on average by 17.3% when linking IoT data. The asymmetric storage efficiency of IoT data according to hash code length is improved by 6.9% on average over existing techniques. Asymmetric storage speed according to the hash code length of the IoT data block was shown to be 10.3% faster on average than existing techniques. Integrity accuracy of IoT data is improved by 18.3% on average over existing techniques.

As the development of IoT technologies has progressed rapidly recently, most IoT data are focused on monitoring and control to process IoT data, but the cost of collecting and linking various IoT data increases, requiring the ability to proactively integrate and analyze collected IoT data so that cloud servers (data centers) can process smartly. In this paper, we propose a blockchain-based IoT big data integrity verification technique to ensure the safety of the Third Party Auditor (TPA), which has a role in auditing the integrity of AIoT data. The proposed technique aims to minimize IoT information loss by multiple blockchain groupings of information and signature keys from IoT devices. The proposed technique allows IoT information to be effectively guaranteed the integrity of AIoT data by linking hash values designated as arbitrary, constant-size blocks with previous blocks in hierarchical chains. The proposed technique performs synchronization using location information between the central server and IoT devices to manage the cost of the integrity of IoT information at low cost. In order to easily control a large number of locations of IoT devices, we perform cross-distributed and blockchain linkage processing under constant rules to improve the load and throughput generated by IoT devices.

As cloud technology advances, devices such as IoT (Internet of Things) are being utilized in various areas ranging from transportation, manufacturing, energy, automation, space, defense, and healthcare. As the number of IoT devices increases, the safety of IoT information, which is vulnerable to cyber attacks, is emerging as an important area of interest in distributed cloud environments. However, integrity techniques are not guaranteed to easily identify the integrity threats and attacks on IoT information operating in the distributed cloud associated with IoT systems and CPS (Cyber-Physical System). In this paper, we propose a blockchain-based integrity verification technique in which large amounts of IoT information processed in distributed cloud environments can be guaranteed integrity in security threats related to IoT systems and CPS. The proposed technique aims to ensure the integrity of IoT information by linking information from IoT devices belonging to subgroups in distributed cloud environments to information from specific non-adjacent IoT devices and blockchain. This is because existing techniques rely on third-party organizations that the data owner can trust to verify the integrity of the data. The proposed technique identifies IoT information by connecting the paths of IoT pre- and subsequent blocks into block chains so that synchronization can be achieved between subgroups in distributed cloud environments. Furthermore, the proposed technique uses probabilistic similarity information between IoT information blocks to react flexibly to subgroups that constitute distributed clouds so that IoT information blocks are not exploited maliciously by third parties. As a result of performance evaluation, the proposed technique averaged 12.3% improvement in integrity processing time over existing techniques depending on blockchain size. Furthermore, the proposed technique has to hash the IoT information that constitutes a subgroup with probability-linked information, validating the integrity of large-capacity IoT information, resulting in an average of 8.8% lower overhead than existing techniques. In addition, the proposed technique has an average improvement of 14.3% in blockchain-based integrity verification accuracy over existing techniques, depending on the hash chain length.

End stage renal disease (ESRD) is the last stage of chronic kidney disease that requires dialysis or a kidney transplant to survive. Many studies reported a higher risk of mortality in ESRD patients compared with patients without ESRD. In this paper, we develop a model to predict postoperative complications, major cardiac event, for patients who underwent any type of surgery. We compare several widely-used machine learning models through experiments with our collected data yellow of size 3220, and achieved F1 score of 0.797 with the random forest model. Based on experimental results, we found that features related to operation (e.g., anesthesia time, operation time, crystal, and colloid) have the biggest impact on model performance, and also found the best combination of features. We believe that this study will allow physicians to provide more appropriate therapy to the ESRD patients by providing information on potential postoperative complications.

Malware detection of non-executables has recently been drawing much attention because ordinary users are vulnerable to such malware. Hangul Word Processor (HWP) is software for editing non-executable text files and is widely used in South Korea. New malware for HWP files continues to appear because of the circumstances between South Korea and North Korea. There have been various studies to solve this problem, but most of them are limited because they require a large amount of effort to define features based on expert knowledge. In this study, we designed a convolutional neural network to detect malware within HWP files. Our proposed model takes a raw byte stream as input and predicts whether it contains malicious actions or not. To incorporate highly variable lengths of HWP byte streams, we propose a new padding method and a spatial pyramid average pooling layer. We experimentally demonstrate that our model is not only effective, but also efficient.

Hypotensive events in the initial stage of anesthesia can cause serious complications in the patients after surgery, which could be fatal. In this study, we intended to predict hypotension after tracheal intubation using machine learning and deep learning techniques after intubation one minute in advance. Meta learning models, such as random forest, extreme gradient boosting (Xgboost), and deep learning models, especially the convolutional neural network (CNN) model and the deep neural network (DNN), were trained to predict hypotension occurring between tracheal intubation and incision, using data from four minutes to one minute before tracheal intubation. Vital records and electronic health records (EHR) for 282 of 319 patients who underwent laparoscopic cholecystectomy from October 2018 to July 2019 were collected. Among the 282 patients, 151 developed post-induction hypotension. Our experiments had two scenarios: using raw vital records and feature engineering on vital records. The experiments on raw data showed that CNN had the best accuracy of 72.63%, followed by random forest (70.32%) and Xgboost (64.6%). The experiments on feature engineering showed that random forest combined with feature selection had the best accuracy of 74.89%, while CNN had a lower accuracy of 68.95% than that of the experiment on raw data. Our study is an extension of previous studies to detect hypotension before intubation with a one-minute advance. To improve accuracy, we built a model using state-of-art algorithms. We found that CNN had a good performance, but that random forest had a better performance when combined with feature selection. In addition, we found that the examination period (data period) is also important.

In this paper, a cylindrical cavity sensor based on microwave resonant theory is proposed to distinguish between various driveability index gasolines under temperature variations. The working principle of the proposed sensor is based on the fact that the change in permittivity of gasoline samples inside cavity sensor will also cause a change in resonant frequency. The proposed sensor has good sensitivity in terms of resonant frequency separation, which enables it to capture the minute permittivity changes and distinguish different gasolines. By using a normal gasoline permittivity of 2.15 and changing sensor dimension parameters, the sensor was designed by high-frequency structure simulator (HFSS). The designed sensor has a resonant frequency of 7.119 GHz for the TM 012 mode with a 19.2 mm radius, a 35 mm height, and one-port coupling probe of 8 mm height. The proposed cylindrical cavity sensor shows advantages of excellent resonant characteristics of small cavity size and small sample amount. To optimize and verify the parameters of the sensor, many experiments have been carried out using HFSS and a vector network analyzer (VNA). Consequently, the proposed sensor is proven to be robust to temperature changes in terms of resonant frequency separation. The minimum frequency separation to distinguish gasoline samples is found to be larger than 29 MHz with reflection coefficients under -11 dB for temperature changes from -35 °C to 0 °C. The consistency of experimental and theoretical results also are presented, which guarantees accuracy of the sensor for the distinction of gasoline.

In this study, a time-dependent corrosion depth estimation method using atmospheric corrosion monitor (ACM) sensor data to evaluate time-dependent corrosion behaviors is proposed. For the time-dependent corrosion depth estimation of uncoated carbon steel and weathering steel, acceleration corrosion tests were conducted in salt-spray corrosion environments and evaluated with a corrosion damage estimation method using ACM sensing data and corrosion loss data of the tested steel specimens. To estimate the time-dependent corrosion depth using corrosion current by an ACM sensor, the relationship between the mean corrosion depth calculated from the weight loss method and the corrosion current was evaluated. The mean corrosion depth was estimated by calculating the corrosion current and evaluating the relationship between the mean corrosion depth and corrosion current during the expected period. From the test and estimation results, the corrosion current demonstrated a good linear correlation with the mean corrosion depth of carbon steel and weathering. The calculated mean corrosion depth is nearly the same as that of the tested specimen, which can be well used to estimate corrosion rate for the uncoated carbon steel and weathering steel.

For the past few decades, action recognition has been attracting many researchers due to its wide use in a variety of applications. Especially with the increasing number of smartphone users, many studies have been conducted using sensors within a smartphone. However, a lot of these studies assume that the users carry the device in specific ways such as by hand, in a pocket, in a bag, etc. This paper investigates the impact of providing an action recognition system with the information of the possession-way of a smartphone, and vice versa. The experimental dataset consists of five possession-ways (hand, backpack, upper-pocket, lower-pocket, and shoulder-bag) and two actions (walking and running) gathered by seven users separately. Various machine learning models including recurrent neural network architectures are employed to explore the relationship between the action recognition and the possession-way recognition. The experimental results show that the assumption of possession-ways of smartphones do affect the performance of action recognition, and vice versa. The results also reveal that a good performance is achieved when both actions and possession-ways are recognized simultaneously.

Nowadays, with wireless sensor networks (WSNs) being widely applied to diverse applications, heterogeneous sensor networks (HSNs), which can simultaneously support multiple sensing tasks in a common sensor field, are being considered as the general form of WSN system deployment. In HSNs, each application generates data packets with a different size, thereby resulting in fairness issues in terms of the network performance. In this paper, we present the design and performance evaluation of a differentiated channel access scheme (abbreviated to DiffCA) to resolve the fairness problem in HSNs. DiffCA achieves fair performance among the application groups by providing each node with an additional backoff counter, whose value varies according to the size of the packets. A mathematical model based on the discrete time Markov chain is presented and is analyzed to measure the performance of DiffCA. The numerical results show that the performance degradation of disadvantaged application groups can be effectively compensated for by DiffCA. Simulation results are given to verify the accuracy of the numerical model.