Performance Comparison of Autoencoder based OFDM Communication System with Wi-Fi

Shiho Oshiro,Takao Toma,Tomohisa Wada International Journal of Computer ScienceNetwork S 2023 International journal of computer science and netw Vol.23 No.5

In this paper, performance of autoencoder based OFDM communication systems is compared with IEEE 802.11a Wireless Lan System (Wi-Fi). The proposed autoencoder based OFDM system is composed of the following steps. First, one sub-carrier's transmitter - channel - receiver system is created by autoencoder. Then learning process of the one sub-carrier autoencoder generates constellation map. Secondly, using the plural sub-carrier autoencoder systems, parallel bundle is configured with inserting IFFT and FFT before and after the channel to configure OFDM system. Finally, the receiver part of the OFDM communication system was updated by re-learning process for adapting channel condition such as multipath channel. For performance comparison, IEEE802.11a and the proposed autoencoder based OFDM system are compared. For channel estimation, Wi-Fi uses initial long preamble to measure channel condition. but Autoencoder needs re-learning process to create an equalizer which compensate a distortion caused by the transmission channel. Therefore, this autoencoder based system has basic advantage to the Wi-Fi system. For the comparison of the system, additive random noise and 2-wave and 4-wave multipaths are assumed in the transmission path with no inter-symbol interference. A simulation was performed to compare the conventional type and the autoencoder. As a result of the simulation, the autoencoder properly generated automatic constellations with QPSK, 16QAM, and 64QAM. In the previous simulation, the received data was relearned, thus the performance was poor, but the performance improved by making the initial value of reception a random number. A function equivalent to an equalizer for multipath channels has been realized in OFDM systems. As a future task, there is not include error correction at this time, we plan to make further improvements by incorporating error correction in the future.

ResNet-Variational AutoEncoder기반 변종 악성코드 패밀리 분류 연구

이영전 ( Young-jeon Lee ),한명묵 ( Myung-mook Han ) 한국인터넷정보학회 2021 인터넷정보학회논문지 Vol.22 No.2

전통적으로 대부분의 악성코드는 도메인 전문가에 의해 추출된 특징 정보를 활용하여 분석되었다. 하지만 이러한 특징 기반의 분석방식은 분석가의 역량에 의존적이며 기존의 악성코드를 변형한 변종 악성코드를 탐지하는 데 한계를 가지고 있다. 본 연구에서는 도메인 전문가의 개입 없이도 변종 악성코드의 패밀리를 분류할 수 있는 ResNet-Variational AutoEncder 기반 변종 악성코드 분류 방법을 제안한다. Variational AutoEncoder 네트워크는 입력값으로 제공되는 훈련 데이터의 학습 과정에서 데이터의 특징을 잘 이해하며 정규 분포 내에서 새로운 데이터를 생성하는 특징을 가지고 있다. 본 연구에서는 Variational AutoEncoder의 학습 과정에서 잠재 변수를 추출을 통해 악성코드의 중요 특징을 추출할 수 있었다. 또한 훈련 데이터의 특징을 더욱 잘 학습하고 학습의 효율성을 높이기 위해 전이 학습을 수행했다. ImageNet Dataset으로 사전학습된 ResNet-152 모델의 학습 파라미터를 Encoder Network의 학습 파라미터로 전이했다. 전이학습을 수행한 ResNet-Variational AutoEncoder의 경우 기존 Variational AutoEncoder에 비해 높은 성능을 보였으며 학습의 효율성을 제공하였다. 한편 변종 악성코드 분류를 위한 방법으로는 앙상블 모델인 Stacking Classifier가 사용되었다. ResNet-VAE 모델의 Encoder Network로 추출한 변종 악성코드 특징 데이터를 바탕으로 Stacking Classifier를 학습한 결과 98.66%의 Accuracy와 98.68의 F1-Score를 얻을 수 있었다. Traditionally, most malicious codes have been analyzed using feature information extracted by domain experts. However, this feature-based analysis method depends on the analyst's capabilities and has limitations in detecting variant malicious codes that have modified existing malicious codes. In this study, we propose a ResNet-Variational AutoEncder-based variant malware classification method that can classify a family of variant malware without domain expert intervention. The Variational AutoEncoder network has the characteristics of creating new data within a normal distribution and understanding the characteristics of the data well in the learning process of training data provided as input values. In this study, important features of malicious code could be extracted by extracting latent variables in the learning process of Variational AutoEncoder. In addition, transfer learning was performed to better learn the characteristics of the training data and increase the efficiency of learning. The learning parameters of the ResNet-152 model pre-trained with the ImageNet Dataset were transferred to the learning parameters of the Encoder Network. The ResNet-Variational AutoEncoder that performed transfer learning showed higher performance than the existing Variational AutoEncoder and provided learning efficiency. Meanwhile, an ensemble model, Stacking Classifier, was used as a method for classifying variant malicious codes. As a result of learning the Stacking Classifier based on the characteristic data of the variant malware extracted by the Encoder Network of the ResNet-VAE model, an accuracy of 98.66% and an F1-Score of 98.68 were obtained.

비지도 학습 기반의 임베딩과 오토인코더를 사용한 침입 탐지 방법

이준우 ( Junwoo Lee ),김강석 ( Kangseok Kim ) 한국정보처리학회 2023 정보처리학회논문지. 소프트웨어 및 데이터 공학 Vol.12 No.8

As advanced cyber threats continue to increase in recent years, it is difficult to detect new types of cyber attacks with existing pattern or signature-based intrusion detection method. Therefore, research on anomaly detection methods using data learning-based artificial intelligence technology is increasing. In addition, supervised learning-based anomaly detection methods are difficult to use in real environments because they require sufficient labeled data for learning. Research on an unsupervised learning-based method that learns from normal data and detects an anomaly by finding a pattern in the data itself has been actively conducted. Therefore, this study aims to extract a latent vector that preserves useful sequence information from sequence log data and develop an anomaly detection learning model using the extracted latent vector. Word2Vec was used to create a dense vector representation corresponding to the characteristics of each sequence, and an unsupervised autoencoder was developed to extract latent vectors from sequence data expressed as dense vectors. The developed autoencoder model is a recurrent neural network GRU (Gated Recurrent Unit) based denoising autoencoder suitable for sequence data, a one-dimensional convolutional neural network-based autoencoder to solve the limited short-term memory problem that GRU can have, and an autoencoder combining GRU and one-dimensional convolution was used. The data used in the experiment is time-series-based NGIDS (Next Generation IDS Dataset) data, and as a result of the experiment, an autoencoder that combines GRU and one-dimensional convolution is better than a model using a GRU-based autoencoder or a one-dimensional convolution-based autoencoder. It was efficient in terms of learning time for extracting useful latent patterns from training data, and showed stable performance with smaller fluctuations in anomaly detection performance.

Finding the best suited autoencoder for reducing model complexity

Kien Mai Ngoc,황명권(Myunggwon Hwang) 한국스마트미디어학회 2021 스마트미디어저널 Vol.10 No.3

Basically, machine learning models use input data to produce results. Sometimes, the input data is too complicated for the models to learn useful patterns. Therefore, feature engineering is a crucial data preprocessing step for constructing a proper feature set to improve the performance of such models. One of the most efficient methods for automating feature engineering is the autoencoder, which transforms the data from its original space into a latent space. However certain factors, including the datasets, the machine learning models, and the number of dimensions of the latent space (denoted by k), should be carefully considered when using the autoencoder. In this study, we design a framework to compare two data preprocessing approaches: with and without autoencoder and to observe the impact of these factors on autoencoder. We then conduct experiments using autoencoders with classifiers on popular datasets. The empirical results provide a perspective regarding the best suited autoencoder for these factors.

Recommendation system using Deep Autoencoder for Tensor data

Jina Park(박진아),Hwan-Seung Yong(용환승) 한국컴퓨터정보학회 2019 韓國컴퓨터情報學會論文誌 Vol.24 No.8

These days, as interest in the recommendation system with deep learning is increasing, a number of related studies to develop a performance for collaborative filtering through autoencoder, a state-of-the-art deep learning neural network architecture has advanced considerably. The purpose of this study is to propose autoencoder which is used by the recommendation system to predict ratings, and we added more hidden layers to the original architecture of autoencoder so that we implemented deep autoencoder with 3 to 5 hidden layers for much deeper architecture. In this paper, therefore we make a comparison between the performance of them. In this research, we use 2-dimensional arrays and 3-dimensional tensor as the input dataset. As a result, we found a correlation between matrix entry of the 3-dimensional dataset such as item-time and user-time and also figured out that deep autoencoder with extra hidden layers generalized even better performance than autoencoder.

공분산을 활용한 콘크리트 오토인코더 기반 비지도 특징 선택 기법 연구

이현세,김민걸,조성인 한국방송∙미디어공학회 2024 방송공학회논문지 Vol.29 No.3

특징 선택은 특징 공학의 한 과정으로 주어진 정형 데이터로부터 유의미한 특징 (feature, column)을 선택하는 것을 목적으로 한다. 딥러닝 기술이 다양한 분야에서 주목할 만한 수행 능력을 보여줌에 따라 특징 선택 분야에서도 딥러닝 기술 기반 연구가 활발히 이루어지고 있다. 본 논문에서는 concrete autoencoder 기반 선택 기법에 주목하였다. Concrete autoencoder란 autoencoder에 concreterandom variable을 적용하여 유의미한 특징을 선택하는 기법이다. 하지만 concrete autoencoder 기법은 특징 선택 시 중복을 허용하고,저차원 벡터 공간 내에서 데이터가 클래스별로 군집화 되지 않는다는 문제가 있다. 따라서 본 논문은 저차원 벡터 공간 내에서 데이터의 특징별 covariance를 고려하는 기법을 제시하고 다양한 데이터를 사용하여 이 기법을 평가한다. 제안하는 방법은 특히 유전적 정보를 담고 있는 바이오 데이터를 사용했을 때 우수한 성능을 보여준다. Feature selection is a feature engineering process that aims to select meaningful features from given structured data. As deeplearning technology shows remarkable performance in various fields, deep learning-based research is also actively studied in thefeature selection field. In particular, the concrete autoencoder method, which selects important features by applying a concreterandom variable to the autoencoder, presented excellent performance in the field of feature selection. However, the concreteautoencoder allows overlap when selecting features and has the problem that data is not clustered by class within alow-dimensional vector space. In this paper, we propose a new feature selection technique that is based on the concreteautoencoder technique and can compensate for the shortcomings of the concrete autoencoder technique. The proposed methodconsiders the covariance of data features in a low-dimensional domain to prevent the redundant feature selection while improvingthe clustering quality of samples within a low-dimensional space. The proposed technique showed superior feature selectionperformance compared to existing techniques, and its superiority is especially evident for biological data containing geneticinformation.

Stacked Autoencoder 기반 악성코드 Feature 정제 기술 연구

김홍비(Hong-bi Kim),이태진(Tae-jin Lee) 한국정보보호학회 2020 정보보호학회논문지 Vol.30 No.4

네트워크의 발전에 따라 악성코드 생성도구가 유포되는 등으로 인해 악성코드의 출현이 기하급수적으로 증가하였으나 기존의 악성코드 탐지 방법을 통한 대응에는 한계가 존재한다. 이러한 상황에 따라 머신러닝 기반의 악성 코드 탐지 방법이 발전하는 추세이며, 본 논문에서는 머신러닝 기반의 악성 코드 탐지를 위해 PE 헤더에서 데이터의 feature를 추출한 후 이를 이용하여 autoencoder를 통해 악성코드를 더 잘 나타내는 feature 및 feature importance를 추출하는 방법에 대한 연구를 진행한다. 본 논문은 악성코드 분석에서 범용적으로 사용되는 PE 파일에서 확인 가능한 DLL/API 등의 정보로 구성된 549개의 feature를 추출하였고 머신러닝의 악성코드 탐지 성능 향상을 위해 추출된 feature를 이용하여 autoencoder를 통해 데이터를 압축적으로 저장함으로써 데이터의 feature를 효과적으로 추출해 우수한 정확도 제공 및 처리 시간을 2배 단축에 성공적임을 증명하였다. 시험 결과는 악성코드 그룹 분류에도 유용함을 보였으며, 향후 SVM과 같은 분류기를 도입하여 더욱 정확한 악성코드 탐지를 위한 연구를 이어갈 예정이다. The advent of malicious code has increased exponentially due to the spread of malicious code generation tools in accordance with the development of the network, but there is a limit to the response through existing malicious code detection methods. According to this situation, a machine learning-based malicious code detection method is evolving, and in this paper, the feature of data is extracted from the PE header for machine-learning-based malicious code detection, and then it is used to automate the malware through autoencoder. Research on how to extract the indicated features and feature importance. In this paper, 549 features composed of information such as DLL/API that can be identified from PE files that are commonly used in malware analysis are extracted, and autoencoder is used through the extracted features to improve the performance of malware detection in machine learning. It was proved to be successful in providing excellent accuracy and reducing the processing time by 2 times by effectively extracting the features of the data by compressively storing the data. The test results have been shown to be useful for classifying malware groups, and in the future, a classifier such as SVM will be introduced to continue research for more accurate malware detection.

LSTM-Autoencoder를 이용한 부유식 풍력터빈 블레이드 피치 시스템의 이상징후 감지

조성필(Seongpil Cho) 항공우주시스템공학회 2024 항공우주시스템공학회지 Vol.18 No.4

본 논문은 부유식 풍력터빈의 블레이드 피치 시스템에서 발생하는 이상을 조기에 감지하기 위한 LSTM-Autoencoder 모델 기반의 이상징후 감지 시스템을 설명한다. 발전소 모니터링 시스템에 활용되는 센서 데이터는 주로 시계열 데이터로 구성되며, LSTM 네트워크는 이러한 시계열 데이터를 분석하기 위해 두 개의 단방향 LSTM 네트워크로 구성된다. 이를 통해 순차 데이터에 숨겨진 장기 의존성을 효과적으로 발견할 수 있다. 한편, 오토인코더 메커니즘은 정상상태 데이터로부터만 학습하여 이상상태를 분류될 수 있기 때문에 이 두 가지 네트워크를 결합하여 시스템에 발생하는 이상징후를 효과적으로 감지할 수 있다. 제안된 프레임워크의 효과를 입증하기 위해 풍력 터빈 모델에서 수집한 실제 다변량 시계열 데이터셋을 적용하였다. LSTM-AE 모델은 높은 이상징후 감지 정확도를 달성하여 우수한 성능을 보였다. This paper presents an anomaly detection system that uses an LSTM-Autoencoder model to identify early-stage anomalies in the blade pitch system of floating wind turbines. The sensor data used in power plant monitoring systems is primarily composed of multivariate time-series data for each component. Comprising two unidirectional LSTM networks, the system skillfully uncovers long-term dependencies hidden within sequential time-series data. The autoencoder mechanism, learning solely from normal state data, effectively classifies abnormal states. Thus, by integrating these two networks, the system can proficiently detect anomalies. To confirm the effectiveness of the proposed framework, a real multivariate time-series dataset collected from a wind turbine model was employed. The LSTM-autoencoder model showed robust performance, achieving high classification accuracy.

오토인코더를 이용한 요인 강화 HAR 모형

박민수,백창룡 한국통계학회 2022 응용통계연구 Vol.35 No.1

Realized volatility is well known to have long memory, strong association with other global financial markets and interdependences among macroeconomic indices such as exchange rate, oil price and interest rates. This paper proposes autoencoder factor-augmented heterogeneous autoregressive (AE-FAHAR) model for realized volatility forecasting. AE-FAHAR incorporates long memory using HAR structure, and exogenous variables into few factors summarized by autoencoder. Autoencoder requires intensive calculation due to its nonlinear structure, however, it is more suitable to summarize complex, possibly nonstationary high-dimensional time series. Our AE-FAHAR model is shown to have smaller out-of-sample forecasting error in empirical analysis. We also discuss pre-training, ensemble in autoencoder to reduce computational cost and estimation errors. 실현 변동성은 강한 종속성을 가짐이 잘 알려져 있으며, 글로벌 금융 시장과 유기적으로 연관이 되어 있을 뿐만 아니라 환율, 유가, 이자율 등의 거시적인 지표와도 밀접한 관계가 있다. 본 논문은 이러한 실현 변동성의 효과적인 예측을 위해서 오토인코더를 이용한 FAHAR (autoencoder factor-augmented heterogeneous autoregressive, AE-FAHAR) 모형을 제안한다. AE-FAHAR 모형은 강한 종속성을 HAR 구조로 반영하고, 외부효과에 대한 영향을 오토인코더를 사용하여 몇 개의 요인으로 추출하여 이를 반영한다. 오토인코더는 비선형 방법으로 요인을 추정하기에 많은 계산 시간이 필요하지만 복잡하고 비정상성을 가질 수 있는 고차원 시계열 자료의 요약에 더 적합하다. 이는 곧 실증 자료 분석을 통해 AE-FAHAR 모형이 예측 오차를 줄임을 확인할 수 있었다. 또한 계산 시간을 줄이고 추정 오차를 줄이기 위해 오토인코더에 사전학습 및 앙상블을 적용하는 등의 방법에 대해서도 논의하였다.

LSTM Autoencoder를 활용한 스마트 온실 데이터 이상 탐지 구현

이정호,임동혁,김태현,김만중,박성진,양오석,백정현 한국지식정보기술학회 2023 한국지식정보기술학회 논문지 Vol.18 No.3

In this paper, LSTM (Long-Short Term Memory), a sequence data processing model, is used as an Autoencoder method that reconstructs input smart farm sensor data composed of encoders and decoders to implement an artificial intelligence model that detects anomaly in smart farm. The LSTM Autoencoder was trained by using the normal data collected from the 60 sensors installed for precise control of the greenhouse environment of the smart farm as input sequence values. The trained model obtained a very low train and validation error in the learning process, and the LSTM Autoencoder model that has finally completed the learning process contains information about normal data distribution in the representation vector, so sequence input is restored based on that information. In this restoration precess, when a sequence input that is out of the normal data distribution is received, it is applied to anomaly detection by taking advantage of the fact that the reconstruction error increases because accurate restoration does not proceed. The error that becomes the threshold value that distinguishes normal and abnormal data can be set using the loss distribution obtained in the learning process, and the error for sequence input can be visualized as an image and used for anomaly detection.