• International Journal of Computer Science & Network Security
• /
• v.23 no.8
• /
• pp.177-189
• /
• 2023

Malware detection is an increasingly important operational focus in cyber security, particularly given the fast pace of such threats (e.g., new malware variants introduced every day). There has been great interest in exploring the use of machine learning techniques in automating and enhancing the effectiveness of malware detection and analysis. In this paper, we present a deep recurrent neural network solution as a stacked Long Short-Term Memory (LSTM) with a pre-training as a regularization method to avoid random network initialization. In our proposal, we use global and short dependencies of the inputs. With pre-training, we avoid random initialization and are able to improve the accuracy and robustness of malware threat hunting. The proposed method speeds up the convergence (in comparison to stacked LSTM) by reducing the length of malware OpCode or bytecode sequences. Hence, the complexity of our final method is reduced. This leads to better accuracy, higher Mattews Correlation Coefficients (MCC), and Area Under the Curve (AUC) in comparison to a standard LSTM with similar detection time. Our proposed method can be applied in real-time malware threat hunting, particularly for safety critical systems such as eHealth or Internet of Military of Things where poor convergence of the model could lead to catastrophic consequences. We evaluate the effectiveness of our proposed method on Windows, Ransomware, Internet of Things (IoT), and Android malware datasets using both static and dynamic analysis. For the IoT malware detection, we also present a comparative summary of the performance on an IoT-specific dataset of our proposed method and the standard stacked LSTM method. More specifically, of our proposed method achieves an accuracy of 99.1% in detecting IoT malware samples, with AUC of 0.985, and MCC of 0.95; thus, outperforming standard LSTM based methods in these key metrics.

• Journal of Internet Computing and Services
• /
• v.21 no.5
• /
• pp.21-29
• /
• 2020

Accurate estimation of human pose relies on backbone method in which its role is to extract feature map. Up to dated, the method of backbone feature extraction is conducted by the plain convolutional neural networks named by CNN and the residual neural networks named by Resnet, both of which have various architectures and performances. The CNN family network such as VGG which is well-known as a multiple stacked hidden layers architecture of deep learning methods, is base and simple while Resnet which is a bottleneck layers architecture yields fewer parameters and outperform. They have achieved inspired results as a backbone network in human pose estimation. However, they were used then followed by different pose estimation networks named by pose parsing module. Therefore, in this paper, we present a comparison between the plain CNN family network (VGG) and bottleneck network (Resnet) as a backbone method in the same pose parsing module. We investigate their performances such as number of parameters, loss score, precision and recall. We experiment them in the bottom-up method of human pose estimation system by adapted the pose parsing module of openpose. Our experimental results show that the backbone method using VGG network outperforms the Resent network with fewer parameter, lower loss score and higher accuracy of precision and recall.

• Journal of the Korean Society of Industry Convergence
• /
• v.27 no.1
• /
• pp.183-193
• /
• 2024

Recurrent Neural Network technology that learns past patterns and predicts future patterns using technology for recognizing and classifying objects is being applied to various industries, economies, and languages. And research for practical use is making a lot of progress. However, research on the application of Recurrent Neural Networks for evaluating and predicting the safety of mechanical structures is insufficient. Accurate detection of external load applied to the outside is required to evaluate the safety of mechanical structures. Learning of Recurrent Neural Networks for this requires a large amount of load data. This study applied the Gated Recurrent Unit technique to examine the possibility of load learning and investigated the possibility of applying a stacked Auto Encoder as a way to secure load data. In addition, the usefulness of learning mechanical loads was analyzed with the Gated Recurrent Unit technique, and the basic setting of related functions and parameters was proposed to secure accuracy in the recognition and prediction of loads.

• The Journal of the Acoustical Society of Korea
• /
• v.35 no.4
• /
• pp.303-309
• /
• 2016

This paper proposes a novel approach to the problem of audio signal clustering using a stacked autoencoder. The proposed stacked autoencoder learns an efficient representation for the input signal, enables clustering constituent signals with similar characteristics, and therefore the original sources can be separated based on the clustering results. STFT (Short-Time Fourier Transform) is performed to extract time-frequency spectrum, and rectangular windows at all the possible locations are used as input values to the autoencoder. The outputs at the middle, encoding layer, are used to cluster the rectangular windows and the original sources are separated by the Wiener filters derived from the clustering results. Source separation experiments were carried out in comparison to the conventional NMF (Non-negative Matrix Factorization), and the estimated sources by the proposed method well represent the characteristics of the orignal sources as shown in the time-frequency representation.

• Journal of Broadcast Engineering
• /
• v.24 no.6
• /
• pp.1113-1121
• /
• 2019

In this paper, we propose the face anti-spoofing method based on combination of luminance and chrominance with convolutional neural networks. The proposed method extracts luminance and chrominance features independently from live and fake faces by using stacked convolutional neural networks and auxiliary networks. Unlike previous methods, an attention module has been adopted to adaptively combine extracted features instead of simply concatenating them. In addition, we propose a new loss function, called the contrast loss, to learn the classifier more efficiently. Specifically, the contrast loss improves the discriminative power of the features by maximizing the distance of the inter-class features while minimizing that of the intra-class features. Experimental results demonstrate that our method achieves the significant improvement for face anti-spoofing compared to existing methods.

• Journal of Korea Multimedia Society
• /
• v.24 no.11
• /
• pp.1552-1559
• /
• 2021

Convolution neural networks (CNNs) show notable performance in image processing and are used as representative core models. CNNs extract and learn features from large amounts of train dataset. In general, it has a structure in which a convolution layer and a fully connected layer are stacked. The core of CNN is the convolution layer. The size of the kernel used for feature extraction and the number that affect the depth of the feature map determine the amount of weight parameters of the CNN that can be learned. These parameters are the main causes of increasing the computational complexity and memory usage of the entire neural network. The most computationally expensive components in CNNs are fully connected and spatial convolution computations. In this paper, we propose a Fourier Convolution Neural Network that performs the operation of the convolution layer in the Fourier domain. We work on modifying and improving the amount of computation by applying the fast fourier transform method. Using the MNIST dataset, the performance was similar to that of the general CNN in terms of accuracy. In terms of operation speed, 7.2% faster operation speed was achieved. An average of 19% faster speed was achieved in experiments using 1024x1024 images and various sizes of kernels.

• Journal of Korea Multimedia Society
• /
• v.23 no.8
• /
• pp.977-985
• /
• 2020

Aiming at the problem of complex feature extraction and low accuracy in human action recognition, this paper proposed a network structure combining batch normalization algorithm with GoogLeNet network model. Applying Batch Normalization idea in the field of image classification to action recognition field, it improved the algorithm by normalizing the network input training sample by mini-batch. For convolutional network, RGB image was the spatial input, and stacked optical flows was the temporal input. Then, it fused the spatio-temporal networks to get the final action recognition result. It trained and evaluated the architecture on the standard video actions benchmarks of UCF101 and HMDB51, which achieved the accuracy of 93.42% and 67.82%. The results show that the improved convolutional neural network has a significant improvement in improving the recognition rate and has obvious advantages in action recognition.

• Proceedings of the Korea Electromagnetic Engineering Society Conference
• /
• 2002.11a
• /
• pp.293-296
• /
• 2002

In this paper, we proposed a beam tilting microstrip patch array antenna for the reception of satellite signals by using low cost copper etched polyester films and foams. The configuration and coupling mechanism of the proposed antenna are similar to the dipole Yagi-Uda antenna. It is composed of 3 layers of polyester films and three layers of foam. In order to prevent unwanted radiation and coupling loss by microstrip feeding networks and parasitic patches, a stacked layer with rectangular slots above the driver patch array is inserted. The 16${\times}$8 element microstrip Patch way antenna is Presented by experimental results. Its beam patterns are affected by many parameters such as sizes of the patches, gap between the patches. characteristics of the substrates, feeding method, etc. Owing to its complexities of various design parameters, both simulation and experiment were performed. The fabricated antenna received DBS signal from KOREASAT 3 by doing nothing but adjusting azimuth direction.

• KIISE Transactions on Computing Practices
• /
• v.23 no.5
• /
• pp.316-321
• /
• 2017

We analyze the online Hangul handwriting recognition problem (HHR) and present solutions based on recurrent neural networks. The solutions are organized according to the three kinds of sequence labeling problem - sequence classifications, segment classification, and temporal classification, with additional consideration of the structural constitution of Hangul characters. We present a stacked gated recurrent unit (GRU) based model as the natural HHR solution in the sequence classification level. The proposed model shows 86.2% accuracy for recognizing 2350 Hangul characters and 98.2% accuracy for recognizing the six types of Hangul characters. We show that the type recognizing model successfully follows the type change as strokes are sequentially written. These results show the potential for RNN models to learn high-level structural information from sequential data.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.10
• /
• pp.1403-1408
• /
• 2021

Recently, FPGA-based AI processors are being studied actively. Deep convolutional neural networks (CNN) are basic computational structures performed by AI processors and require a very large amount of multiplication. Considering that the multiplication coefficients used in CNN inference operation are all constants and that an FPGA is easy to design a multiplier tailored to a specific coefficient, this paper proposes a methodology to optimize the multiplier. The method utilizes 2's complement and distributive law to minimize the number of bits with a value of 1 in a multiplication coefficient, and thereby reduces the number of required stacked adders. As a result of applying this method to the actual example of implementing CNN in FPGA, the logic usage is reduced by up to 30.2% and the propagation delay is also reduced by up to 22%. Even when implemented with an ASIC chip, the hardware area is reduced by up to 35% and the delay is reduced by up to 19.2%.