• Review of KIISC
• /
• v.30 no.6
• /
• pp.107-115
• /
• 2020

다양한 응용분야에서 심층신경망 기반의 학습 모델이 앞 다투어 이용됨에 따라 인공지능의 설명 가능한 동작 원리 해석과, 추론이 갖는 불확실성에 관한 분석 또한 심도 있게 연구되고 있다. 이에 심층신경망 기반 기계학습 모델의 취약성이 수면 위로 드러났으며, 이러한 취약성을 이용하여 악의적으로 모델을 공격함으로써 오동작을 유도하고자 하는 시도가 다방면으로 이루어짐에 의해 학습 모델의 강건함 보장은 보안 분야에서의 쟁점으로 부각되고 있다. 모델 추론의 입력으로 이용되는 이미지에 교란값을 추가함으로써 심층신경망의 오분류를 발생시키는 임의의 변형된 이미지를 적대적 사례라 정의하며, 본 논문에서는 최근 인공지능 및 컴퓨터비전 분야에서 이루어지고 있는 이미지 기반 적대적 사례의 생성 기법에 대하여 논한다.

• Korean Journal of Social Welfare Studies
• /
• v.47 no.1
• /
• pp.151-178
• /
• 2016

The primary goal of the current study was to investigate the longitudinal relationship between household income, main caregiver depression, and youth aggression from the interactionist perspective. The data were derived by combining the 2006, 2009 and 2012 survey waves from the Korean Welfare Panel Study. This data set covered the full span of adolescence from elementary to high school. The study utilized 561 families as the final sample and conducted autoregressive cross-lagged analysis. As a result, the early income status, main caregiver depression and youth aggression were likely maintained over time. Second, the results provided support for a reciprocal relationship between income and main caregiver depression. On the other hand, the reciprocal relationship between main caregiver depression and youth aggression was not found in the current study. Finally, the mediating effect of main caregiver depression between income and youth aggression was not found in the present study. In conclusion, the results of this study support the interactionist model in that the association between family income and main caregiver depression involves reciprocity and mutual influence across time. These findings have major implications for policy and interventions in regards to low-income families.

• Journal of Digital Contents Society
• /
• v.12 no.2
• /
• pp.165-176
• /
• 2011

Recently, digital rights management (DRM) related researches are widely spreading with prosperity of IT industries. The DRM technology protects proprietor of copyright by preventing mischanneling and illegal copy. In this paper, we propose a new DRM model that has an enhanced and efficient protocol based on certificate using smart card. The proposed model overcomes weaknesses of WCDRM model and has following additional advantages: first, copy protection is enhanced by hiding user's specific information from attacker by storing the information within smart card; second, server load for contents encryption is reduced by making clear protocols among author, distributer, certificate authority, and users; third, offline user authentication is guaranteed by combining partial secret values in media players and smart card. Exposure of core information also is minimized by storing them in smart card. In addition, we show that the proposed system is more secure than WCDRM model by comparing various factors of anonymous attackers.

• The Journal of the Acoustical Society of Korea
• /
• v.33 no.6
• /
• pp.400-406
• /
• 2014

This paper describes a robust watermark decoding model using a SVM(Support Vector Machine). First, the embedding process is performed inversely for a watermarked signal. And then the watermark is extracted using the proposed model. For SVM training of the proposed model, data are generated that are watermarks extracted from sounds containing watermarks by four different embedding schemes. BER(Bit Error Rate) values of the data are utilized to determine a threshold value employed to create training set. To evaluate the robustness, 14 attacks selected in StirMark, SMDI and STEP2000 benchmarking are applied. Consequently, the proposed model outperformed previous method in PSNR(Peak Signal to Noise Ratio) and BER. It is noticeable that the proposed method achieves BER 1% below in the case of PSNR greater than 10 dB.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.11 no.2
• /
• pp.13-26
• /
• 2001

This paper presents the network security modeling methodology and simulation using the hierarchical and modular modeling and simulation framework. Recently, Howard and Amoroso developed the cause-effect model of the cyber attack, defense, and consequences, Cohen has been proposed the simplified network security simulation methodology using the cause-effect model, however, it is not clear that it can support more complex network security model and also the model-based cyber attack simulation. To deal with this problem, we have adopted the hierarchical and modular modeling and simulation environment so called the System Entity Structure/Model Base (SES/MB) framework which integrates the dynamic-based formalism of simulation with the symbolic formalism of AI. Several simulation tests performed on sample network system verify the soundness of our method.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.33 no.4
• /
• pp.621-631
• /
• 2023

Adversarial training improves the robustness of deep neural networks for adversarial examples. However, the previous adversarial training method focuses only on the adversarial loss function, ignoring that even a small perturbation of the input layer causes a significant change in the hidden layer features. Consequently, the accuracy of a defended model is reduced for various untrained situations such as clean samples or other attack techniques. Therefore, an architectural perspective is necessary to improve feature representation power to solve this problem. In this paper, we apply an attention module that generates an attention map of an input image to a general model and performs PGD adversarial training upon the augmented model. In our experiments on the CIFAR-10 dataset, the attention augmented model showed higher accuracy than the general model regardless of the network structure. In particular, the robust accuracy of our approach was consistently higher for various attacks such as PGD, FGSM, and BIM and more powerful adversaries. By visualizing the attention map, we further confirmed that the attention module extracts features of the correct class even for adversarial examples.

• Journal of Digital Convergence
• /
• v.11 no.11
• /
• pp.675-682
• /
• 2013

This paper tries to solve the problems which previous methods have the model using smart card for protecting digital contents. This study provides a contents distribution model to protect the rights of author, distributor, and user as well as user's information by using technologies such as cryptography, DRM(Digital Right Management), access control, etc. The proposed system is evaluated as the most safety model compared with previous methods because it not only solves the problems which the previous methods have, but also protects four type of risks such as use of contents which other mobile devices download, the attack on the key to decode the message, the attack on leaking the contents, and the internal attack such as an illegal reproduction.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.13 no.5
• /
• pp.113-127
• /
• 2003

WTLS as secure protocol of WAP makes TLS that is used in wireless Intemet protocol for TCP security be appropriate for wireless environments. And purpose of WTLS is to provide safe and efficient services. WTLS protocol consists of 4 protocols(Handshake, ChangeCipherSpec, Alert, Application Data etc.). In this papers we analyze properties of Handshake protocol and procedures of establishing master secret in detail. And then we analyze securities against several attacker models with them for a basis. Also we propose new Handshake protocol that is secure against active attacker model and can provide various security services.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.16 no.2
• /
• pp.71-80
• /
• 2006

To guarantee secure communication in wireless sensor networks, secret keys should be securely established between sensor nodes. Recently, a simple key distribution scheme has been proposed for pair-wise key establishment in sensor networks by Anderson, Chan, and Perrig. They defined a practical attack model for non-critical commodity sensor networks. Unfortunately, the scheme is vulnerable under their attack model. In this paper, we describe the vulnerability in their scheme and propose a modified one. Our scheme is secure under their attack model and the security of our scheme is proved. Furthermore, our scheme does not require additional communication overhead nor additional infrastructure to load potential keys into sensor nodes.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.33 no.3
• /
• pp.449-458
• /
• 2023

Artificial Intelligence is providing convenience in various fields using big data and deep learning technologies. However, deep learning technology is highly vulnerable to adversarial examples, which can cause misclassification of classification models. This study proposes a method to detect and purification various adversarial attacks using StarGAN. The proposed method trains a StarGAN model with added Categorical Entropy loss using adversarial examples generated by various attack methods to enable the Discriminator to detect adversarial examples and the Generator to purification them. Experimental results using the CIFAR-10 dataset showed an average detection performance of approximately 68.77%, an average purification performance of approximately 72.20%, and an average defense performance of approximately 93.11% derived from restoration and detection performance.