• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.5
• /
• pp.817-826
• /
• 2022

AI technology is being successfully introduced in many fields, and models deployed as a service are deployed with black box environment that does not expose the model's information to protect intellectual property rights and data. In a black box environment, attackers try to steal data or parameters used during training by using model output. This paper proposes a method of inferring the type of model to directly find out the composition of layer of the target model, based on the fact that there is no attack to infer the information about the type of model from the deep learning model. With ResNet, VGGNet, AlexNet, and simple convolutional neural network models trained with MNIST datasets, we show that the types of models can be inferred using the output values in the gray box and black box environments of the each model. In addition, we inferred the type of model with approximately 83% accuracy in the black box environment if we train the big and small relationship feature that proposed in this paper together, the results show that the model type can be infrerred even in situations where only partial information is given to attackers, not raw probability vectors.

• Korean Journal of Cognitive Science
• /
• v.28 no.2
• /
• pp.111-131
• /
• 2017

The connectionist model is one approach to studying language processing from a computational perspective. And building a representation in the connectionist model study is just as important as making the structure of the model in that it determines the level of learning and performance of the model. The connectionist model has been constructed in two different ways: localist representation and distributed representation. However, the localist representation used in the previous studies had limitations in that the unit of the output layer having a rare target activation value is inactivated, and the past distributed representation has the limitation of difficulty in confirming the result by the opacity of the displayed information. This has been a limitation of the overall connection model study. In this paper, we present a new method to induce distributed representation with local representation using abstraction of information, which is a feature of restricted Boltzmann machine, with respect to the limitation of such representation of the past. As a result, our proposed method effectively solves the problem of conventional representation by using the method of information compression and inverse transformation of distributed representation into local representation.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.17 no.10
• /
• pp.719-724
• /
• 2016

This paper describes the performance improvement of an inrush current limiter to prevent damage or malfunctions in power converters due to the inrush current. When the power converter of military radar is operated, the circuit breaker of the power converter is often activated because the overcurrent flows through the circuit breaker of the power converter. Therefore, this study performed a cause analysis of the problem, which is a larger current flow than the intended current(250A). The operation principle of an inrush current limiter and SCR (Silicon Controlled Rectifier) used in the inrush current limiter was analyzed. As a result, the overcurrent flow through the circuit breaker was found to be due to dv/dt triggering of SCR. Based on cause analysis, this paper proposes a technique by adding the resistor in front of the SCR to prevent an unnecessary inrush current. Finally, the effectiveness of the improvement was verified by measuring the output current in the inrush current limiter. The power converter equipped with the improved inrush current limiter operated for more than 1 year without the circuit breaker of the power converter being activated.

• Korean Chemical Engineering Research
• /
• v.60 no.1
• /
• pp.93-99
• /
• 2022

With the rapid development of ultra-small and wearable device technology, continuous electricity supply without spatiotemporal limitations for driving electronic devices is required. Accordingly, Triboelectric nanogenerator (TENG), which utilizes static electricity generated by the contact and separation of two different materials, is being used as a means of effectively harvesting various types of energy dispersed without complex processes and designs due to its simple principle. However, to apply the TENG to real life, it is necessary to increase the electrical output. In addition, stable generation of electrical output, as well as increase in electrical output, is a task to be solved for the commercialization of TENG. In this study, we proposed a method to not only improve the output of TENG but also to stably represent the improved output. This was solved by using the contact layer, which is one of the components of TENG, as an electret for improved output and stability. The utilized electret was manufactured by sequentially performing corona charging-thermal annealing-corona charging on the Fluorinated ethylene propylene (FEP) film. Electric charges artificially injected due to corona charging enter a deep trap through the thermal annealing, so an electret that minimizes charge escape was fabricated and used in TENG. The output performance of the manufactured electret was verified by measuring the voltage output of the TENG in vertical contact separation mode, and the electret treated to the corona charging showed an output voltage 12 times higher than that of the pristine FEP film. The time and humidity stability of the electret was confirmed by measuring the output voltage of the TENG after exposing the electret to a general external environment and extreme humidity environment. In addition, it was shown that it can be applied to real-life by operating the LED by applying an electret to the clap-TENG with the motif of clap.

• Aerospace Engineering and Technology
• /
• v.11 no.2
• /
• pp.39-48
• /
• 2012

In case a measurement output of gyro sensor is an accumulated angle counts, it is usually provided as a binary bit counter which is allowed to roll-over at its maximum or minimum value. And it is a well known fact that the roll-over behavior restricts the measurement range of the processed sensor output below the actual measurable range of sensor hardware itself. In this study, a conventional sensor data processing method for a gyro with an accumulated angle output is introduced. And also, an improved method which can extend the processed output range over the conventional one is proposed. It is also derived that the increased range depends on the variation speed of a input signal. Finally, the derived equations and the performance of the proposed algorithm are verified using a computer simulation.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.35S no.6
• /
• pp.1-8
• /
• 1998

This paper presents an asynchronous DS/CDMA multiuser detector, which is a two stage, symbol-by-symbol scheme consisting of conventional detectors followed by linear decorrelating detectors. The conventional detector first makes temporal decisions and the detected symbols are delayed by one symbol period to be used for the selection of decorrelating bases in the subsequent decorrelaing detection stage. It also employs a bank of early-late correlators in place of a bank of single correlators taking the small offset of chip timing asynchronism into account. The proposed detector requires only the coarse knowledge of relative time delays of interfering users and is suitable for digital implementation. To verify the detector performance, the analytical BER performance will be given and compared with the simulation results for BPSK DS/CDMA signals in AWGN channel. While the performance of the proposed detector will be analyzed for time-limited signal, the simulation is carried out for both the time-limited and band-limited signals. As can be seen in the simulation results, the proposed scheme shows good results.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.16 no.1
• /
• pp.103-108
• /
• 2016

In communication systems the solution of the problem of maximizing the mutual information between the input and output of a channel composed of several subchannels under total power constraint has a waterfilling structure. OFDM and MIMO can be decomposed into parallel subchannels with CSI. Waterfilling solves the problem of optimal power allocation to these subchannels to achieve the rate approaching the channel capacity under total power constraint. In waterfilling, more power is alloted to good channels(high SNR) and less or no power to bad channels to increase the rate of good channels, resulting in channel capacity. Waterfilling finds the exact water level satisfying the power constraint employing an iterative algorithm to estimate and update the water level. In this process computation of partial sums of inverse of square of subchannel gain is repeatedly required. In this paper we reduced the computation time of waterfilling algorithm by replacing the partial sum computation with reference to an array which contains the precomputed partial sums in initialization phase.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.12 no.6
• /
• pp.1121-1126
• /
• 2017

Various studies such as INS(: Inertial Navigation System) are conducting to estimate the position of UAV, because the GPS information of UAV is at risk like the GPS jamming. The position estimation using DoA and RTT are used to apply many radar systems, and that process can be applied in datalink of UAV. The general monopulse feed in UAV datalink is Multi-horn, because of the wide BW(: Band Width) and frequency range. And it needs wide SNR range of tracking because of the limited transmit power of airborne unit. The estimation error of position increase at low SNR, and the DoA is valid in only 3dB beam width but high SNR causes false of mainlobe detection because of large sidelobe. In this paper, We propose the method to achieve higher accuracy of DoA estimation on low SNR and review some idea that able to detect mainlobe.

• Journal of IKEEE
• /
• v.24 no.1
• /
• pp.225-231
• /
• 2020

In this paper, we propose a method to accelerate BWN based on FPGA with limited resources for embedded system. Because of the limited number of logic elements available, a single computing unit capable of handling Conv-layer, FC-layer of various sizes must be designed and reused. Also, if the input feature map can not be parallel processed at one time, the output must be calculated by reading the inputs several times. Since the number of available BRAM modules is limited, the number of data bits in the BWN accelerator must be minimized. The image classification processing time of the BWN accelerator is superior when compared with a embedded CPU and is faster than a desktop PC and 50% slower than a GPU system. Since the BWN accelerator uses a slow clock of 50MHz, it can be seen that the BWN accelerator is advantageous in performance versus power.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.38A no.9
• /
• pp.744-749
• /
• 2013

We propose soft user data input on LDPC codes with parity encoded by the (1, 7) run length limited (RLL) code for perpendicular magnetic recording channel. The user data are encoded by maximum transition run (MTR) (3;11) code. In order to minimize the loss of code rate, the (1, 7) RLL code only encode the parity of LDPC. Also, to increase performance, we propose only user data part applied soft output Viterbi algorithm (SOVA). The performance using the SOVA showed good performance lower than 26 dB. In contrast, it showed worse performance high than 26 dB. This is because of incorrect soft information by high jitter noise and two different input types for LDPC decoder.