於虛擬化環境下考量誘捕系統及動態拓樸調整以達到攻擊者成功機率最小化之有效網路建置與防禦策略

Abstract

虛擬化在企業發展過程中扮演重要的角色。透過該技術，使用者可不受現有的硬體架構和地域限制，彈性的進行運算與儲存資源，但同時也使得虛擬化環境的資訊安全議題更加複雜。身為服務提供者兼防禦者，在服務廣大的合法使用者之餘，也須面臨各式各樣的攻擊者與日新月異的攻擊手法，因此如何在惡意攻擊下最大化系統存活度成為一個極度重要的議題。除採取適當的資源配置策略外，因現實世界中攻擊者大多對於欲攻擊的目標僅擁有「不完全資訊」，並未完全掌握防禦者所使用的防禦機制，故攻防過程中防禦者可透過誘捕系統和動態拓樸調整達到欺騙攻擊者與消耗其資源的目的。在維持一定服務品質水準的前提下降低核心節點被攻克的機率，提升整體網路系統的存活度。 在本論文中，我們將攻防情境轉化成一個數學規劃問題，用以描述網路系統被攻擊者攻克的機率，並提出一套以鬆弛觀念與蒙地卡羅法為基礎的解題方法，結合數學規劃法與模擬，處理更貼近真實情況的問題。在模擬的過程中，藉由每次評估所獲得的相關資訊，逐步調整並推導出最適當的修正方向。最終得出能使網路系統被攻克機率最小化之防禦資源配置與相對應的防禦策略。

Virtualization plays an important role in the enterprise development. Through this technology, users can access computing power and storage resource flexibly without the limitation of hardware framework and geography. However, it also raises the complexity of information security in the virtualization environment. As a service provider, we serve numerous legitimate users and strive against the variety of attackers with the diversity of attack tactics simultaneously. Therefore, how to maximize the survivability of network system under malicious attack becomes an extremely notable subject. Since most attackers only have “incomplete information” of the targeted system in the real world and only have a little knowledge about defense mechanisms, the defender can distract attackers and waste their budget by deception techniques and dynamic topology reconfiguration. Moreover, the defender should decrease the compromised probability of core nodes and maintain the specific Quality of Service level at the same time. In this thesis, we model the attack-defense scenario as a mathematical programming problem that describes attackers’ success probability and propose a solution approach which combines the mathematical programming and simulation. Based on the concept of relaxation and Monte Carlo simulation, the scale of solvable problem is extended. In the process of simulation, we can gradually improve the quality of solution and conclude the most appropriate revised direction via the information gathered from each evaluation. Finally, the experiment result comprising the defense resource allocation and corresponding defense strategies for the defender to minimize the compromised probability of network system.