[논문]k-익명화 알고리즘에서 기계학습 기반의 k값 예측 기법 실험 및 구현

장성봉

doi:10.3745/ktccs.2020.9.1.9

k-익명화 알고리즘에서 기계학습 기반의 k값 예측 기법 실험 및 구현
Experiment and Implementation of a Machine-Learning Based k-Value Prediction Scheme in a k-Anonymity Algorithm 원문보기

정보처리학회논문지. KIPS transactions on computer and communication systems 컴퓨터 및 통신 시스템, v.9 no.1, 2020년, pp.9 - 16

(금오공과대학교 IT융복합공학과) , 장성봉 (금오공과대학교 산학협력단)

초록
AI-Helper

빅 데이터를 연구 목적으로 제3자에게 배포할 때 프라이버시 정보를 보호하기 위해서 k-익명화 기법이 널리 사용되어 왔다. k-익명화 기법을 적용할 때, 해결 해야할 어려운 문제 중의 하나는 최적의 k값을 결정하는 것이다. 현재는 대부분 전문가의 직관에 근거하여 수동으로 결정되고 있다. 이러한 방식은 익명화의 성능을 떨어뜨리고 시간과 비용을 많이 낭비하게 만든다. 이러한 문제점을 해결하기 위해서 기계학습 기반의 k값 결정방식을 제안한다. 본 논문에서는 제안된 아이디어를 실제로 적용한 구현 및 실험 내용에 대해서 서술 한다. 실험에서는 심층 신경망을 구현하여 훈련하고 테스트를 수행 하였다. 실험결과 훈련 에러는 전형적인 신경망에서 보여지는 패턴을 나타냈으며, 테스트 실험에서는 훈련에러에서 나타나는 패턴과는 다른 패턴을 보여주고 있다. 제안된 방식의 장점은 k값 결정시 시간과 비용을 줄일 수 있다는 장점이 있다.

Abstract ▼ AI-Helper

The k-anonymity scheme has been widely used to protect private information when Big Data are distributed to a third party for research purposes. When the scheme is applied, an optimal k value determination is one of difficult problems to be resolved because many factors should be considered. Currently, the determination has been done almost manually by human experts with their intuition. This leads to degrade performance of the anonymization, and it takes much time and cost for them to do a task. To overcome this problem, a simple idea has been proposed that is based on machine learning. This paper describes implementations and experiments to realize the proposed idea. In thi work, a deep neural network (DNN) is implemented using tensorflow libraries, and it is trained and tested using input dataset. The experiment results show that a trend of training errors follows a typical pattern in DNN, but for validation errors, our model represents a different pattern from one shown in typical training process. The advantage of the proposed approach is that it can reduce time and cost for experts to determine k value because it can be done semi-automatically.

주제어

표/그림 (13)

표 Table 1. An Example of k-anonymity with k=2
그림 Fig. 1. An Example of Weights Updates in a Artificial Neural Network
그림 Fig. 2. A Comparison of Conventional Scheme and Proposed Scheme
그림 Fig. 3. Architecture of Neural Networks Model Defined in a Proposed Approach
표 Table 2. Permitted Level of Loss Ratio
표 Table 3. An Example of Published Data
표 Table 4. Sample Data Set
그림 Fig. 4. Determination Process of an Optimal Neural Network Model
그림 Fig. 5. An Example of Tenforflow Codes Implemented in a Experiment
표 Table 5. Candidate Neural Network Model for k determination
그림 Fig. 6. Training Errors Distribution in an Experiment
그림 Fig. 7. Validation Errors Distribution in an Experiment
표 Table 6. RMSE Errors and Processing Time for Each Model

AI 본문요약
AI-Helper

* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.

가설 설정

The concept is based on the observation that if there are no critical information in the source data, we don’t need to anonymize the data because there are no information to protect.

제안 방법

The concept are described in works of [9-11]. In that work, the authors presented a basic scheme for machine learning with artificial intelligence. The scheme asserts that a machine can learn something through long-time training.
The purpose of the method was to enhance the privacy of Big Data while guaranteeing data utility. The approach consisted of six steps: Tuple Partitioning, Optional Column Generalization (Map), Attribute Partitioning (Reduce), and Suppression Slicing. Mohamed Nassar [06] proposed a Query language for data anonymization.
[05] presented a method that is based on slicing adapted from the suppression technique. The purpose of the method was to enhance the privacy of Big Data while guaranteeing data utility. The approach consisted of six steps: Tuple Partitioning, Optional Column Generalization (Map), Attribute Partitioning (Reduce), and Suppression Slicing.

대상 데이터

Various models has been evaluated from the aspect of RMSE errors and processing time, and one model has been chosen as target model. The chosen model has been trained and tested using a dataset that has three features. The experiment results shows that the trend of the training errors follows a typical pattern in ANN, which is an exponential decrease.

이론/모형

In this work, the ANN model has been defined and implemented using Tensorflow libraries to estimate a k value in a k-anonymity algorithm. Various models has been evaluated from the aspect of RMSE errors and processing time, and one model has been chosen as target model.
The last component was a data anonymizer which was responsible for anonymizing the data transferred from the request handler. It implemented and contained algorithms such as k-anonymity and l-diversity. They have implemented a prototype using Java and Hadoop server, and they conducted an experiment to evaluate their approach.
Xuyun Zhang [08] pointed out a local recording problem when k-anonymization applied to Big Data. To solve the problem, they presented a proximity privacy model. The solution considered semantic similarity among sensitive attribute values during big data anonymization.

성능/효과

The solution considered semantic similarity among sensitive attribute values during big data anonymization. The advantage of the scheme was that it was easy to make the model more scalable. Also, they implemented the algorithms with MapReduce for parallel execution.
The chosen model has been trained and tested using a dataset that has three features. The experiment results shows that the trend of the training errors follows a typical pattern in ANN, which is an exponential decrease. When the number of iteration reaches about 10,000, the error value becomes stable.

참고문헌 (18)

L. M. Kumbayoni and S.-B. Jang, "A Determination of k-value Based on Artificial Neural Network in k-anonymization," Proceedings of International Conference on Convergence Research, Vol.4, No.3, pp.437-440, 2018.
L. Sweeney, "k-anonymity: a model for protecting privacy," International Journal on Uncertainty, Fuzziness and Knowledge based Systems, pp.557-570, 2002.
B. C. M. Fung, K. Wang, R. Chen, and P. S. Yu, "Privacy-Preserving Data Publishing: A Survey of Recent Developments," ACM Computing Survey, Vol.42, No.4, pp.1-53, 2010.
I. Roy, S. T. Setty, A Kilzer, V Shmatikov, and E Witchel, "Airavat: Security and Privacy for Mapreduce," Proceedings of Seventh USENIX Conference on Networked Systems Design and Implementation, pp.297-312, 2010.
N. Elanshekhar and R. Shedge, "An effective anonymization technique of big data using suppression slicing method," 2017 International Conference on Energy, Communication, Data Analytics and Soft Computing (ICECDS), pp.2500- 2504, 2017,
M. Nassar, A. A. R. Orabi, M. Doha, and B. Al Bouna, "An SQL-like query tool for data anonymization and outsourcing," 2015 International Conference on Cyber Situational Awareness, Data Analytics and Assessment (CyberSA), pp.1-3, 2015.
R. Matsunaga, I. Ricarte, T. Basso, and R. Moraes, "Towards an Ontology-Based definition of Data Anonymization Policy for Cloud Computing and Big Data," 2017 47th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshops (DSN-W), pp.75-82, 2017.
X. Zhang, W. Dou, J. Pei, S. Nepal, C. Yang, C. Liu, and J. Chen, "Proximity-Aware Local-Recoding Anonymization with MapReduce for Scalable Big Data Privacy Preservation in Cloud," IEEE Transactions on Computers, Vol.64, No.8, pp.2293-2307, 2015.

상세보기
A. A. William, "On the Efficiency of Learning Machines," IEEE Transactions on Systems Science and Cybernetics, Vol.3, No.2, pp.111-116, 1967.

상세보기
V. F. Nicholas, "Some New Approaches to Machine Learning," IEEE Transactions on Systems Science and Cybernetics, Vol.5, No.3, pp.173-182, 1969.

상세보기
H. C. Anderson, "Neural network machines", IEEE Potentials, Vol.8, No.1, pp.13-16, 1989.

상세보기
J. Nagumo and A. Noda, "A learning method for system identification," IEEE Transactions on Automatic Control, Vol.12, No.3, pp.282-287, 1967.

상세보기
W. T. Illingworth, "Beginner's guide to neural networks," IEEE Aerospace and Electronic Systems Magazine, Vol.4, No.9, pp.234-241, 1989.
D. Burshtein, "Long-term attraction in higher order neural networks," IEEE Transactions on Neural Networks, Vol.9, No.1, pp.42-50, 1998.

상세보기
P. D. Wasserman and T. Schwartz, "Neural networks. II. What are they and why is everybody so interested in them now?," IEEE Expert, Vol.3, No.1, pp.10-15, 1988.

상세보기
P. Yugowati, M. Shaou-Gang, and W. Hui-Ming, "Supervised learning approaches and feature selection-a case study in diabetes," International Journal of Data Analysis Techniques and Strategies, Vol. 5, No.3, 2013, pp. 323-337.

상세보기
J.-X. Peng, L. Kang, and W. I. George, "A New Jacobian Matrix for Optimal Learning of Single-Layer Neural Networks," IEEE Transactions on Neural Networks, Vol.19, No.1, pp.119-129, 2008.

상세보기
C. Gustavo, B. C. Antoni, J. M. Pedro, and V. Nuno, "Supervised Learning of Semantic Classes for Image Annotation and Retrieval," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.29, No.3, pp.394-410, 2007.

상세보기

저자의 다른 논문 :

표제어: PCR

동의어: Packet Collision Rate

용어 설명 출처 목록 (6)

용어 설명: PCR은 세균 특이성이 있는 primer를 이용하여 적은 수의 세균이 있을지라도 쉽게 검출할 수 있는 유용한 방법이며, 이를 이용하여 구강 내 치면세균막이나 타액에서 직접 세균을 검출할 수 있게 되었다[8].

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증