氏名　Zaw Zaw Aung
学位の種類　博士（工学）
学位記番号　博甲第554号
学位授与の日付　平成22年8月31日
学位論文題目　Modeling for Managing Risk of Operational and Economic Interdependency among Japanese Critical Infrastructures　（日本の重要インフラ間のオペレーション上および経済的な相互依存性に係わるリスクマネジメントモデルの構築）
論文審査委員
　主査　教授　淺井　達雄
　副査　教授　大里　有生
　副査　教授　三上　喜貴
　副査　准教授　五島　洋行
　副査　名古屋工業大学教授　渡辺　研司

• 論文目次
• 論文要旨
• 論文審査の結果の要旨

［平成22（2010）年度博士論文題名一覧］ ［博士論文題名一覧］に戻る．

Table of Contents
Table of Contents　p.i
List of Figures　p.iii
Abstract　p.v
Chapter 1.　Introductioon　p.1
　1.1.　Background　p.1
　1.2.　Study Objectives　p.1
　1.3.　The conceptual view of structural dependency　p.2
　1.4.　Basic Assumptions　p.3
　1.5.　Data Sources　p.6
　1.6.　Scope and Limitations　p.6
　1.7.　Outline of the Thesis　p.7
Chapter 2.　Critical Infrastructure Protection　p.9
　2.1.　Introduction　p.9
　2.2.　International Movement on CIP　p.10
　2.3.　Japan Critical Infrastructure Protection　p.11
Chapter 3.　Risk Management Framework for Critical Infrastructure Protection　p.17
　3.1.　The Framework　p.17
　3.2.　Risk Assessment　p.20
　3.3.　Consequence Asessment　p.22
　3.4.　Interdependency of Critical Infrastructures　p.23
　3.5.　Modeling Total Economic Impact of Extreme Event　p.26
Chapter 4.　Comparative Study of Operational vs. Economic dependency　p.33
　4.1.　Introduction　p.33
　4.2.　Data Sources　p.33
　4.3.　Influence Matrices and Total Requirement Tables　p.36
　4.4.　Influence Driving and Influence Receving　p.42
　4.5.　Net Influence and Strength of Relation　p.43
Chapter 5.　Inoperability Input-output Model （IIM）　p.45
　5.1.　Introduction　p.45
　5.2.　Leontief Economic Model　p.49
　5.3.　Inoperability Input-output Model　p.51
　5.3.1　Physical-based IIM　p.52
　5.3.2　Demand-reduction IIM　p.53
　5.4　Developing Demand-reduction IIM for Japan　p.55
　5.5　Assumptions and Limitations　p.69
Chapter 6.　Bayesian Networks （BN）　p.71
　6.1.　Introduction　p.71
　6.2.　Casual Modeling Concepts　p.71
　6.3.　Bayes'Theorem　p.72
　6.4.　Fundamental Rule and Bayes' Rule　p.73
　6.5.　Theory of Bayesian Networks　p.73
　6.6.　Directed Acyclic Graph （DAG）　p.74
　6.7.　Nodes, Network and Structural Validation　p.74
　6.8.　Criteria for Using Probabilistic Network　p.77
Chapter 7.　The Proposed Bayesian-IMM Mixed Model　p.79
　7.1.　Background　p.79
　7.2.　Problem Description and the Approach　p.80
　7.3.　The Basic Assumptions for the Proposed Model　p.82
　7.4.　The Conceptual View of Proposed Model　p.82
Chapter 8.　The Case Study　p.85
　8.1.　Introduction　p.85
　8.2.　Specific Objectives of this Case Study　p.86
　8.3.　Data Sources　p.86
　8.4.　Construction of a Bayesian Network for Inoperability Propagation　p.88
　8.4.1.　Structural Model Building　p.88
　8.4.2.　Quantifying the Network　p.89
　8.5.　Case study on future earthquake underneath Tokyo　p.94
　8.6.　Results and Applications　p.102
Chapter 9.　Concluding remarks, Limitations and Future Works　p.105
　9.1.　Concluding remarks　p.105
　9.2.　Limitations and Future Works　p.107
References　p.111
List of Publications　p.117
International Conferences　p.118
Local Conferences and Workshops　p.119
Acknowledgements　p.121

　Critical infrastructures （CI） are formed by business entities providing highly irreplaceable services and are essential for people’s social lives and economic activities. In order to improve their ability to provide goods and services efficiently and cost effectively, infrastructures must interact with one another at various levels. However, these interactions increase infrastructure dependencies, rendering the entire system extremely complex and prone to "domino failures". The task of analyzing the impact of infrastructure outages （goods and services） on the overall system is formidable.
　Traditional systems analysis methodologies yield results of limited value because they fail to consider mutual dependency （interdependency） phenomena.

　This study focuses on analyzing the effect of interdependency existing in critical infrastructure by combining Inoperability Input-output model （IIM） and Bayesian probabilistic networks. IIM was developed by Yacov Haimes and his colleagues and it is based on the well-known theory of market equilibrium developed by Nobel laureate, Wassily Leontief. The model assumes that there exists a direct correlation between national economic input-output data and operability/inoperability of economic sectors. However, this correlation represents a very crude approximation.
The nation’s input-output data represent economic dependencies among sector which is insignificant in some cases. Japan has defined ten sectors as critical infrastructures- namely Telecommunication, Government and Administrative Services, Finance, Civil Aviation, Railways, Logistics, Electricity, Gas, Medical Services, Water works and are evidently and mostly utility services sectors and they take very insignificant figure in input-output tables. However, the higher degree of physical and functional dependencies can be observed among them.

　The proposed model use Bayesian causal network as a buffer between initial perturbations and IIM for allowing flexible adjustment and risk management interventions while solving the above mentioned problem. Data mining approach is used to extract dependency data from publically published reports especially based on interdependency study report published by National Institute of Land and Infrastructure Management （NILIM） of Japan. The various form of dependency existing in Japan’s infrastructure are analyzed comparing NILIM’s influence matrices and dependency matrix generated from Year 2000 National Input-output tables published by Statistics Bureau of Japan. Demand-reduction Inoperability Input-output Model is developed using 104 sector Input-output tables and parametric analysis of interdependency impact on the whole nation economy is done by setting 10% to 50% perturbation constantly to ten critical infrastructures. These analyses underline which sectors/CI are mostly dependent on operability/inoperability of other sectors and how the level of service disruption is propagated throughout the economic sectors.

　The conceptual view of structural dependency among Japanese critical infrastructures and economic sectors is as discuss below. Critical infrastructures are located at the core of economic sectors and they are tightly coupled each other in functional, physical and share resources. Therefore, any service disruption at one or more of critical infrastructure has greater impact to other CI. Non-CI sectors rely heavily on services provided by one or more critical infrastructures. One economic sector may not rely on certain CI （say it, CI-A）. However, it rely on another CI （say it CI-B） that is tightly coupled with CI-A. In such case, the economic sector can suffer some impact if CI-A has certain service disruption. For instance, a bank may have backup power supply. However, if electricity disruption occurs at the region and telecommunication hubs and switches cannot operate, the bank will suffer indirect impact. We have limited resources to study all kind of interdependency issues in each and every economic sector. Therefore, the approach of this study is to model interdependency of critical infrastructures as much detail as possible and use it as a buffer for estimating potential losses and consequences.

　Bayesian network for ten critical infrastructure is constructed based on dependency information published by National Information Security Center （NISC） and quantified the network judging from NILIM survey matrix and parametric analysis of demand reduction IIM. To evaluate the developed Model, a case study on future earthquake underneath Tokyo Metro Area is conducted. The Report published by Tokyo Metropolitan Disaster Management Council provides detail estimate of intensity and potential service disruptions in 23 areas in Tokyo Metro due to predicted M7.3 earthquake. The case study uses its data as initial perturbation for the model and estimate sector-wise inoperability as well as economic loss of 104 economic sectors. Additionally, it discusses how to use the model results to present the impact of the earthquake in other dimension apart from economic loss ? such as lifeline impact, Nation （economic） security impact, etc. The developed model is found to be effective to calculate more accurate economic as well as other dimensional impacts. It provides better refined picture of impact from certain system outage and assists in risk and disaster management decision making.

　本論文は、「Modeling for Managing Risk of Operational and Economic Interdependency among Japanese Critical Infrastructures（日本の重要インフラ間のオペレーション上および経済的な相互依存性に係わるリスクマネジメントモデルの構築）」と題し、8章より構成されている。

　第1章「Introduction」では、本論文のベースとなる構造的な依存性についての概念を説明し、研究に用いた分析アプローチの整理しながら、本論文全体の流れを述べている。

　第2章「Critical Infrastructure Protection（CIP）」では、重要インフラ防護（CIP）の必要性と、CIPに係る国際研究の動向を説明し、日本におけるCIPの現状を俯瞰している。

　第3章「Risk Management Framework for CIP」では、CIPに必要なリスクマネジメントの枠組みを解説した上で、リスク評価、事案発生時の結果についての評価の詳細を述べている。また、重要インフラ間の相互依存性解析に基づいた大規模災害時の経済的影響予測のモデルを示している。

　第4章「Comparative Study of Operational vs. Economic dependency」では、本論文で用いた産業連関表データの概要を述べると同時に、オペレーション上の依存性と経済的な依存性の比較分析を行っている。

　第5章「Inoperabiliy Input-output Model（IIM）」では、本論文における相互依存性解析の基幹のひとつとなるIIMモデルについて、その起源となるレオンチェフ理論とその限界も含めて議論を展開している。

　第6章「Bayesian Networks」では、本論文における相互依存性解析のもうひとつの基盤となる、ベイジアン・ネットワーク・モデルを解説し、その中心となる有向非巡回グラフ（DAG）を用いた依存性解析および確率推論の有効性に関する議論を展開している。

　第7章「The Case Study」では、内閣官房、国土交通省、東京都が公開している、重要インフラの相互依存性に係る統計・分析および、大規模地震の被害想定に基づき、本論文で議論を展開した重要インフラ間の相互依存分析の有効性と限界を示している。

　第8章「Conclusion」では、上記の各章で得られた知見に基づき、IIMとベイジアン・ネットワーク・モデルを併用した、重要インフラ相互依存性解析の有効性を検証し結論を導くとともに、本研究課題の今後の展開についても言及している。

　よって、本論文は工学上及び工業上貢献するところが大きく、博士（工学）の学位論文として十分な価値を有するものと認める。