• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.3
• /
• pp.1100-1118
• /
• 2021

In a wide range of ML applications, the training data contains privacy-sensitive information that should be kept secure. Training the ML systems by privacy-sensitive data makes the ML model inherent to the data. As the structure of the model has been fine-tuned by training data, the model can be abused for accessing the data by the estimation in a reverse process called model inversion attack (MIA). Although, MIA has been applied to shallow neural network models of recognizers in literature and its threat in privacy violation has been approved, in the case of a deep learning (DL) model, its efficiency was under question. It was due to the complexity of a DL model structure, big number of DL model parameters, the huge size of training data, big number of registered users to a DL model and thereof big number of class labels. This research work first analyses the possibility of MIA on a deep learning model of a recognition system, namely a face recognizer. Second, despite the conventional MIA under the white box scenario of having partial access to the users' non-sensitive information in addition to the model structure, the MIA is implemented on a deep face recognition system by just having the model structure and parameters but not any user information. In this aspect, it is under a semi-white box scenario or in other words a gray-box scenario. The experimental results in targeting five registered users of a CNN-based face recognition system approve the possibility of regeneration of users' face images even for a deep model by MIA under a gray box scenario. Although, for some images the evaluation recognition score is low and the generated images are not easily recognizable, but for some other images the score is high and facial features of the targeted identities are observable. The objective and subjective evaluations demonstrate that privacy cyber-attack by MIA on a deep recognition system not only is feasible but also is a serious threat with increasing alert state in the future as there is considerable potential for integration more advanced ML techniques to MIA.

• Journal of Korean Society of Environmental Engineers
• /
• v.39 no.4
• /
• pp.220-228
• /
• 2017

Since the fine particulate matters occurred from mainly combustion in industry and road transport effect to human respiratory health, the interest and importance are getting increased. In 2013, the World Health Organization (WHO) concluded that outdoor air pollution is carcinogenic to humans, with the particulate matter component ($PM_{10}$ and $PM_{2.5}$) of air pollution most closely associated with increased cancer incidence, especially cancer of the lung. Therefore, many researches have been studied in the quantification and data development of fine particulate matters. Currently, the Ministry of Environment and cities/regions are developing the fine particulate matter data and air emission information. Particularly just $PM_{10}$ and $PM_{2.5}$ data is used in the fine particulate matters warning and alert. The data of NOx, SOx, $NH_3$, which have the particulate matter formation potential are not well considered. Also, the researches related with particulate matter formation potential and respiratory effects by industrial sectors and cities/regions are not conducted well. Therefore, the purpose of this study is to evaluate and calculate particulate matter formation potential and respiratory effects in 11 industrial sectors and cities using NOx, SOx, $PM_{10}$, $NH_3$ data (developed by Ministry of Environment and National Institute of Environmental Research) in 2001 and 2013. The results of this study will be provided the particulate matter formation potential and respiratory effects and will be used for future the fine particulate matter researches.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.9
• /
• pp.1898-1905
• /
• 2012

This paper presents an On-board Smart Device (OSD) for moving vehicle, based on a smooth integration of Android-based devices and a Micro-control Unit (MCU). The MCU is used for the acquisition and transmission of various vehicle-borne data. The OSD has threefold functions: Record, Report and Alarm. Based on these RRA functions, the OSD is basically a safety and convenience oriented smart device, where it facilitates alert services such as accident report and rescue as well as alarm for the status of vehicle. In addition, voice activated interface is developed for the convenience of users. Vehicle data can also be uploaded to a remote server for further access and data manipulation. Therefore, unlike conventional blackboxes, the developed OSD lends itself to a user-friendly smart device for vehicle safety: It basically stores monitoring images in driving plus vehicle data collection. Also, it reports on accident and enables subsequent rescue operation. The developed OSD can thus be considered an essential safety smart device equipped with comprehensive wireless data service, image transfer and voice activated interface.

• Journal of Internet Computing and Services
• /
• v.17 no.6
• /
• pp.33-39
• /
• 2016

With recent advances in sensor technology, CMOS-based semiconductor devices and networking protocol, the areas for application of wireless sensor networks greatly expanded and diversified. Such diversification of uses for wireless sensor networks creates a multitude of beneficial possibilities for several industries. In the application of wireless sensor networks for monitoring systems' data transmission process from the sensor node to the sink node, transmission through multi-hop paths have been used. Also mobile sink techniques have been applied. However, high energy costs, unbalanced energy consumption of nodes and time gaps between the measured data values and the actual value have created a need for advancement. Therefore, this thesis proposes a new model which alleviates these problems. To reduce the communication costs due to frequent data exchange, a State Prediction Model has been developed to predict the situation of the peripheral node using a geographic autocorrelation of sensor nodes constituting the wireless sensor networks. Also, a Risk Analysis Model has developed to quickly alert the monitoring system of any fatal abnormalities when they occur. Simulation results have shown, in the case of applying the State Prediction Model, errors were smaller than otherwise. When the Risk Analysis Model is applied, the data transfer latency was reduced. The results of this study are expected to be utilized in any efficient communication method for wireless sensor network monitoring systems where all nodes are able to identify their geographic location.

• Publications of The Korean Astronomical Society
• /
• v.27 no.4
• /
• pp.285-286
• /
• 2012

We combine data from two all-sky surveys, the Swift/Burst Alert Telescope 22 Month Source Catalog and the AKARI Point Source Catalogue, in order to study the connection between the hard X-ray (> 10 keV) and infrared (IR) properties of local active galactic nuclei (AGN). We find two photometric diagnostics are useful for source classification: one is the X-ray luminosity vs. IR color diagram, in which type 1 radio-loud AGN are well isolated from other AGN. The second one uses the X-ray vs. IR color-color diagram as a redshift-independent indicator for identifying Compton-thick (CT) AGN. Importantly, CT AGN and starburst galaxies in composite systems can also be separated in this plane based upon their hard X-ray fluxes and dust temperatures. This diagram may be useful as a new indicator to classify objects in new surveys such as with WISE and NuSTAR.

• Science of Emotion and Sensibility
• /
• v.17 no.4
• /
• pp.51-60
• /
• 2014

This double-blind study, as a human experiment of nonlocality, investigated the effects of senders' intention on the central nervous system of a distant human receiver and it explored the roles that motivation might have in modulating these effects. Whole brain activity was measured in the receiver whom was asked to relax in a distant room for 16 minutes; the sending person directed intention of oneness toward the receiver during repeated variable-second epochs separated by variable-second non-intention epochs. The total length of intention epochs and that of nonintention epochs were balanced. Eighteen sessions were conducted. In 9 of those sessions, the sender was the receiver's lover. In another 9 of those sessions, the sender was just acquainted with the receiver before the session. The receiver's whole brain activity recorded during the intention epochs were compared with the same measures recorded during the nonintention epochs used as controls. The statistical difference between the intentions versus controls across 18 sessions was examined by paired-t test. In addition, subgroup analysis for the 9 couple sessions and 9 non-couple sessions were additionally examined by the same test. The effect of distant intentionality decreased slow waves or increased EEG fast waves mainly in frontal regions, and increased EEG coherence during the intention epochs. The effects was not statistically significant after Bonferroni correction, but the couple sessions combined showed the largest effect followed by all sessions combined. Non-couple sessions combined showed the smallest effect. The changes in EEG power mean that receiver participants became more alert during the intention epochs and the change in EEG coherence might be evidence of coherent heart influence on EEG activity. Planned comparison with specific hypothesis testing for the suggested changes in this study have to be followed for an evidence of electroencephalographic changes in isolated subjects for the distant mental intention.

• Journal of Korean Society for Atmospheric Environment
• /
• v.15 no.5
• /
• pp.545-554
• /
• 1999

We would like to develop a short-term model to predict the time-related concentration of ozone whose reaction mechanism is complex. The paper targets Seoul where an ozone alert system has recently been employed. In order to develop a short-term prediction model for ozone, we suggested the Ozone Peak Indicator(OPI), an equivalent of the potential daily maximum ozone concentration, with precursors being the only limiting factor, and we calculated the Ozone Peak Indicarot as OPI={$rac{(O_3)_{max}cdot(H_{eH})_{max}(Rad)_{max}$ to preclude the influence of mixing height and solar radiation on the daily maximum ozone concentration. The OPI on the day of the prediction is to be calcultated by using the relation between OPI and the initial value of precursors. The basic prediction formula for time-related ozone concentration was established as $O_3(1)={(OPI)cdot Rad(t-2)H_{eH}}$, using the OPI, solar radiation two hours before prediction and mixing height. We developed, along with the basic formula for predicting photochemical oxidants, "SEOM"(Seoul Empirical Oxidants Model), a Fortran program that helps predict solar radiation and mixing height needed in the prediction of ozone pollution. When this model was applied to Seoul and an analysis of the correlation between the observed and the predicted ozone concentrations was made through SEOM, there appeared a very high correlation, with a coefficient of 0.815. SEOM can be described as a short-term prediction model for ozone concentration in large cities that takes into account the initial values of precursors, and changes in solar radiation and mixing height. SEOM can reflect the local characteristics of a particular and region can yield relatively good prediction results by a simple data input process.t process.

• Journal of Korean Neurosurgical Society
• /
• v.42 no.1
• /
• pp.1-5
• /
• 2007

Objective : We studied whether frontal skull base fracture has an impact on the occurrence and recovery of anosmia and/or ageusia following frontal traumatic brain injury (TBI). Methods : Between May 2003 and April 2005, 102 consecutive patients who had hemorrhage or contusion on the frontal lobe base were conservatively treated. Relevant clinical and radiographic data were collected, and assessment of impaired smell and taste sensation were also surveyed up to at least 12 months post-injury. Results : Among 102 patients, anosmia was noted in 22 (21.6%), of whom 10 had ageusia at a mean 4.4 days after trauma. Bilateral frontal lobe injuries were noted in 20 of 22 patients with anosmia and in all 10 patients with ageusia. Frontal skull base fracture was noted in 41 patients, of whom 9 (21.4%) had anosmia and 4 (9.5%) had ageusia. There was no statistical difference in the occurrence of anosmia and ageusia between patients with or without fracture. Of the 22 patients with anosmia, recovery from anosmia occurred in nine (40.9%) at the interval of 6 to 24 months after trauma, of whom six had frontal skull base fracture and three were not associated with fracture. Recovery of anosmia was significantly higher in patients without fracture than those with fracture (p<0.05). Recovery from ageusia occurred in only two of 10 patients at the interval of 18 to 20 months after trauma and was not eminent in patients without fracture. Conclusion : One should be alert and seek possibile occurrence of the anosmia and/or ageusia following frontal TBI. It is suggested that recovery is quite less likely if such patients have fractures on the frontal base, and these patients should wait for at least 6 to 18 months to anticipate such recovery if there is no injury to the central olfactory structures.

• KIISE Transactions on Computing Practices
• /
• v.23 no.11
• /
• pp.641-649
• /
• 2017

As Internet of Things devices become pervasive, services improve to better assess the context and to alert other users to deal with emergencies. Such services use Internet of Things devices to detect the context around the user and promptly notify public institutions, hospitals or other parental users in emergencies. Most of these systems analyze an event when the value of the device is unchanged for a period of time or if it detects an abnormal value. However, just monitoring sensor values makes it difficult to accurately understand the context surrounding a user. Also if the device is inactive, it can not identify the context or provide services again. However, understanding the user requirements, services provided through other devices, information sent to other users lets, appropriate actions be taken. This paper, proposes a device search method and system based on a context-aware model that includes user demands. The proposed system analyzes the user's context and demands by using data collected from the internet of things devices. If user devices are inactive, they can recognize other devices by searching for other devices and providing services to users again. Through the proposed method, the user-centric services are provided. This method also analyzes and responds to requirements in various emergencies.

• Journal of Korean Academy of Nursing
• /
• v.24 no.3
• /
• pp.345-363
• /
• 1994

This study attempted to identify important problems of the factors related to the indwelling catheter with closed drainage and to enhance recognization of Urinary Tract Infection management to protect patients with indwelling catheter from UTI. We selected the 58 patients with indwelling catheter with closed drainage from md-surg. intensive care units, and general neurological and neurosurgical ward, in one hospital in Pusan, Korea from May 30 to October 15, 1993. Patients were examined by urinary analysis before inserting catheter and then by urine cultures three times. The obtained data was analized using an SPSS /PC$^{+}$ The statistical analyses employed here was the frequency, percentage and x$^2$-tests. 1. Analyses for Indwelling Catheter induced UTI and infection rates were found to be presented in patients related to the factors of being on antibiotic drugs (x$^2$=14.105, p=.000) and the duration of the indwelling catheter(x$^2$=58.0, p=.000), both of which were found to affect the UTI. 2. UTI rates according to the duration of the in-dwelling catheter were high 8.6％ for the 2nd day, 27.6 for the 4th day, and 36.2% for the 7th day, and as the duration for the indwelling catheter was longer, UTI rates were higher. 3. The UTI rates according to the duration of the indwelling catheter related re factors, on the 7th day, were high-52.0% for women, 37.0% for more than 50 years in ages, 30.0% for patients under sur-gery, 33.3% for altered mental states outsides the alert, 76.7% for patients not given antibiotic drugs, 36.4% for less than 3500cc of fluids intakes, it was found that these affected to increase the UTI. 4. Causative organisms of UTI were found ; E.Coli appears the most than all other isolated organisms in women and Klebsiella and Pseudomonas in men. For antibiotic susceptibility of causative organisms, all causative organisms were found to be susceptible to Sul -cefoperazone, Trim-prim, and unasyn and Pseudomonas mostly net susceptible.e.