Soo Young SHIN Hong Seong PARK Wook Hyun KWON
In this paper, the packet error rate (PER) of IEEE 802.15.4 under the interference of a saturated IEEE 802.11b network is evaluated using an analytic model when IEEE 802.15.4 and IEEE 802.11b coexist. The PER is obtained from the bit error rate (BER) and the collision time, where the BER is obtained from the signal-to-interference-plus-noise ratio. The analytic results are validated using simulations.
Zhonghua QUAN Soohee HAN Wook Hyun KWON
We propose a stability-guaranteed horizon size (SgHS) for stabilizing receding horizon control (RHC). It is shown that the proposed SgHS can be represented explicitly in terms of the known parameters of the given system model and is independent of the terminal weighting matrix in the cost function. The proposed SgHS is validated via a numerical example.
Hyun KWON Hyunsoo YOON Ki-Woong PARK
Malicious attackers on the Internet use automated attack programs to disrupt the use of services via mass spamming, unnecessary bulletin boarding, and account creation. Completely automated public turing test to tell computers and humans apart (CAPTCHA) is used as a security solution to prevent such automated attacks. CAPTCHA is a system that determines whether the user is a machine or a person by providing distorted letters, voices, and images that only humans can understand. However, new attack techniques such as optical character recognition (OCR) and deep neural networks (DNN) have been used to bypass CAPTCHA. In this paper, we propose a method to generate CAPTCHA images by using the fast-gradient sign method (FGSM), iterative FGSM (I-FGSM), and the DeepFool method. We used the CAPTCHA image provided by python as the dataset and Tensorflow as the machine learning library. The experimental results show that the CAPTCHA image generated via FGSM, I-FGSM, and DeepFool methods exhibits a 0% recognition rate with ε=0.15 for FGSM, a 0% recognition rate with α=0.1 with 50 iterations for I-FGSM, and a 45% recognition rate with 150 iterations for the DeepFool method.
Hyun KWON Hyunsoo YOON Ki-Woong PARK
We propose a multi-targeted backdoor that misleads different models to different classes. The method trains multiple models with data that include specific triggers that will be misclassified by different models into different classes. For example, an attacker can use a single multi-targeted backdoor sample to make model A recognize it as a stop sign, model B as a left-turn sign, model C as a right-turn sign, and model D as a U-turn sign. We used MNIST and Fashion-MNIST as experimental datasets and Tensorflow as a machine learning library. Experimental results show that the proposed method with a trigger can cause misclassification as different classes by different models with a 100% attack success rate on MNIST and Fashion-MNIST while maintaining the 97.18% and 91.1% accuracy, respectively, on data without a trigger.
Chung-Ho SHIN Chul-Hyun KWON Su-Yeon KIM Sang-Hui PARK
A novel caption text detection and recognition algorithm using the temporal nature of video is proposed in this paper. A text registration technique is used to locate the temporal and spatial positions of captions in video from the accumulated frame difference information. Experimental results show that the proposed method is effective and robust. Also, a high processing speed is achieved since no time consuming operation is included.
Jae-Min LEE Soo Hee HAN Hong Seong PARK Wook Hyun KWON
In this paper, a refined analytic model is presented for the IEEE 802.11 distributed coordination function (DCF) in a time-varying channel environment. In the proposed model, the channel is modelled using a finite-state Markov (FSM) chain. The saturation throughput and average packet delay are analyzed from the proposed model. It is shown using OPNETTM and UltraSANTM simulations that the proposed model accurately predicts the performance of the IEEE 802.11 DCF.
Kamrok LEE Jae Yeol HA Hong Seong PARK Wook Hyun KWON
This paper analyzes the throughput and the optimal announcement traffic indication message (ATIM) window of the IEEE 802.11 Distributed Coordination Function (DCF) in the power saving mode. An analytical model based on Markov chain model is proposed to express the throughput and the optimal ATIM window in a mathematical form; it is validated by the simulation. The optimal ATIM window size is obtained to maximize the throughput and minimize the power consumption while solving the fairness problem.
Hyun KWON Yongchul KIM Ki-Woong PARK Hyunsoo YOON Daeseon CHOI
Deep neural networks (DNNs) are widely used in many applications such as image, voice, and pattern recognition. However, it has recently been shown that a DNN can be vulnerable to a small distortion in images that humans cannot distinguish. This type of attack is known as an adversarial example and is a significant threat to deep learning systems. The unknown-target-oriented generalized adversarial example that can deceive most DNN classifiers is even more threatening. We propose a generalized adversarial example attack method that can effectively attack unknown classifiers by using a hierarchical ensemble method. Our proposed scheme creates advanced ensemble adversarial examples to achieve reasonable attack success rates for unknown classifiers. Our experiment results show that the proposed method can achieve attack success rates for an unknown classifier of up to 9.25% and 18.94% higher on MNIST data and 4.1% and 13% higher on CIFAR10 data compared with the previous ensemble method and the conventional baseline method, respectively.
Naehyuck GHANG Jaehyun PARK Wook Hyun KWON
This paper proposes a hardware architecture of programmable controller based on Petri nets. The suggested architecture achieves sufficiently rapid processing even as demands on PCs become increasingly more complex. The architecture's speed and efficiency are derived from an automatic and dynamic super scalar computing capability that executes bit instructions and data handling instructions simultaneously without preprocessing, due to the properties of Petri nets. Specific characteristics for both architectural memory-based implementation of Petri nets and evolution algorithms are suggested and classified by the net structure. Analysis of the suggested architectures and effects on performance are also given with mathematical formulas and a computer simulation.
MyungSeon RYOU HongSeong PARK SooHee HAN WookHyun KWON
This letter discusses the prediction of the time-varying bit error rate (BER) for a transmitting channel using recent transmissions and retransmissions. Depending on the predicted BER, we propose a maximum frame size control to improve the goodput in wireless networks. It is shown, using simulation, that when the maximum frame size is controlled relative to the time-varying BER the goodput of the network is improved.
Bin-Chul IHM Dong-Jo PARK Young-Hyun KWON
We propose a new intelligent blind source separation algorithm for the mixture of sub-Gaussian and super-Gaussian sources. The algorithm consists of an update equation of the separating matrix and an adjustment equation of nonlinear functions. To verify the validity of the proposed algorithm, we compare the proposed algorithm with extant methods.
Seung-Rae LEE Wook Hyun KWON Koeng-Mo SUNG
In this paper, the previous definition of the Reverse Jacket matrix (RJM) is revised and generalized. In particular, it is shown that the inverse of the RJM can be obtained easily by a constructive approach similar to that used for the RJM itself. As new results, some useful properties of RJMs, such as commutativity and the Hamiltonian symmetry appearing in half the blocks of a RJM, are shown, and also 1-D fast Reverse Jacket transform (FRJT) is presented. The algorithm of the FRJT is remarkably efficient than that of the center-weighted Hadamard transform (CWHT). The FRJT is extended in terms of the Kronecker products of the Hadamard matrix. The 1-D FRJT is applied to the discrete Fourier transform (DFT) with order 4, and the N-point DFT can be expressed in terms of matrix decomposition by using 4 4 FRJT.
Hyeok Gi PARK Hong-ju MOON Wook Hyun KWON
In this paper a cyclic place-timed controlled marked graph (PTCMG), which is an extended class of a cyclic controlled marked graph (CMG), is presented as a model of discrete event systems (DESs). In a PTCMG, time constraints are attached to places instead of transitions. The time required for a marked place to be marked again is represented in terms of time constraints attached to places. The times required for an unmarked place to be marked under various controls, are calculated. The necessary and sufficient condition for a current marking to be in the admissible marking set with respect to the given forbidden condition is provided, as is the necessary and sufficient condition for a current marking to be out of the admissible marking set with respect to the forbidden condition in one transition. A maximally permissive state feedback control is synthesized in a PTCMG to guarantee a larger admissible marking set than a CMG for most forbidden state problems. Practical applications are illustrated for a railroad crossing problem and an automated guided vehicle (AGV) coordination problem in a flexible manufacturing facility.
Ohmin KWON Hyun KWON Hyunsoo YOON
We propose a rootkit installation method inside a GPU kernel execution process which works through GPU context manipulation. In GPU-based applications such as deep learning computations and cryptographic operations, the proposed method uses the feature by which the execution flow of the GPU kernel obeys the GPU context information in GPU memory. The proposed method consists of two key ideas. The first is GPU code manipulation, which is able to hijack the execution flow of the original GPU kernel to execute an injected payload without affecting the original GPU computation result. The second is a self-page-table update execution during which the GPU kernel updates its page table to access any location in system memory. After the installation, the malicious payload is executed only in the GPU kernel, and any no evidence remains in system memory. Thus, it cannot be detected by conventional rootkit detection methods.
Deep neural networks show good performance in image recognition, speech recognition, and pattern analysis. However, deep neural networks also have weaknesses, one of which is vulnerability to poisoning attacks. A poisoning attack reduces the accuracy of a model by training the model on malicious data. A number of studies have been conducted on such poisoning attacks. The existing type of poisoning attack causes misrecognition by one classifier. In certain situations, however, it is necessary for multiple models to misrecognize certain data as different specific classes. For example, if there are enemy autonomous vehicles A, B, and C, a poisoning attack could mislead A to turn to the left, B to stop, and C to turn to the right simply by using a traffic sign. In this paper, we propose a multi-targeted poisoning attack method that causes each of several models to misrecognize certain data as a different target class. This study used MNIST and CIFAR10 as datasets and Tensorflow as a machine learning library. The experimental results show that the proposed scheme has a 100% average attack success rate on MNIST and CIFAR10 when malicious data accounting for 5% of the training dataset have been used for training.
ChoonKi AHN SooHee HAN WookHyun KWON
This letter presents parametric uncertainty bounds (PUBs) for stabilizing receding horizon H∞ control (RHHC). The proposed PUBs are obtained easily by solving convex optimization problems represented by linear matrix inequalities (LMIs). We show, by numerical example, that the RHHC can guarantee a H∞ norm bound for a larger class of uncertain systems than conventional infinite horizon H∞ control (IHHC).
Let V(φ) be a shift invariant subspace of L2(R) with a Riesz or frame generator φ(t). We take φ(t) suitably so that the regular sampling expansion : f(t) = f(n)S(t-n) holds on V(φ). We then find conditions on the generator φ(t) and various bounds of the perturbation {δ n }n∈Z under which an irregular sampling expansion: f(t) = f(n+ δn)Sn(t) holds on V(φ). Some illustrating examples are also provided.
Chul-Hyun KWON Doo-Jin HAN Hyun-Sool KIM Myung-Ho LEE Sang-Hui PARK
Shot transition detection is a core technology in video browsing, indexing systems and information retrieval. In this paper we propose a dissolve detection algorithm using the characteristics of edge in MPEG compressed video. Using the intensity change information of edge pixels obtained by Sobel edge detector, we detect the location of a dissolve and its precise duration. We also present a new reliable method to eliminate the false dissolves. The proposed algorithm is tested in various types of videos, and the experimental results show that the proposed algorithm is effective and robust.
Bin-Chul IHM Dong-Jo PARK Young-Hyun KWON
We propose a blind source separation algorithm for the mixture of finite alphabet sources where sensors are less than sources. The algorithm consists of an update equation of an estimated mixing matrix and enumeration of the inferred sources. We present the bound of a step size for the stability of the algorithm and two methods of assignment of the initial point of the estimated mixing matrix. Simulation results verify the proposed algorithm.
Sinuk KANG Kil Hyun KWON Dae Gwan LEE
We present a multi-channel sampling expansion for signals with selectively tiled band-region. From this we derive an oversampling expansion for any bandpass signal, and show that any finitely many missing samples from two-channel oversampling expansion can always be uniquely recovered. In addition, we find a sufficient condition under which some infinitely many missing samples can be recovered. Numerical stability of the recovery process is also discussed in terms of the oversampling rate and distribution of the missing samples.