於雲端運算環境中對惡意攻擊之損害最小化資源分配策略

Abstract

近年來，雲端運算已經對企業的基礎設施架構產生了巨大的影響，雲端廠商和企業的合作變得至關重要，隨著越來越多應用在雲端上的開發，網路安全的問題也跟著浮現，雲端廠商和企業都需要承擔責任來防禦惡意攻擊。因此，我們提供了一個數學模型，代表兩方對於利益的博弈關係。在網路攻防的情境下，一個資源配置的策略來抵禦惡意攻擊，兩方競爭者的利益相互衝突，防禦者想要最大化虛擬機的預期存活度，攻擊者想要最小化虛擬機的預期存活度，競爭成功函數是一個我們用來衡量競爭者成功的機率函數。 拉格朗日乘數法和拉格朗日放鬆法是兩個我們用來解決這個數學模型的方法，在實驗中，我們比較了不同方法的優缺點，像是暴力法，近似最佳法，和啟發式演算法，不同方法之間的功效，得到的結果是合理且符合我們的預期。

Cloud computing has made a huge impact on enterprise infrastructures. The cooperation of cloud providers and the firm is crucial. As more applications are being developed, network security issues emerge. Both of them should take the responsibilities and defense the malicious attacks. Thus, we develop a general mathematical model as a two-player max-min game. A resource allocation strategy is presented in a network defense and attack scenario. The interests of the two players contradict. One wants to maximize the expected value of VMs while the other wants to minimize it. Contest Success Function is a function we use to measure the compromise probability. Lagrange multiplier and Lagrangian relaxation are the methods we applied to solve the problem. In our experiment, we compare advantages and disadvantages for different approaches including brute force, near-optimal and heuristic methods. The effectiveness of different approaches is also presented. The result we gain is reasonable and corresponds to our expectation.