雲端計算可以透過虛擬化技術共享軟硬體資源，但是使用者在使用虛擬化平台時可能會面臨額外的的安全威脅。共存攻擊是指攻擊者利用共享基礎設備的特性，來攻擊共存在同一實體機上的其他虛擬機。2017年Abazari等人提出了雲端共存攻擊多目標回應系統，考量到虛擬機共存時間，以最小成本和最小威脅為目標來回應共存攻擊。但我們發現實際上該系統的回應時間過長。因此在本論文中，我們使用機器學習來訓練入侵回應系統，我們使用Ridge Regression演算法，並進行一系列實驗證明模型的效果，實驗中也比較了我們的模型和Abazari等人的模型。實驗結果顯示，我們的模型具有 2000 倍加速比的效率提升，同時獲得 87.9% 高準確度的解答，在回應策略與時間效能取得平衡。 With cloud computing, we can share hardware and software resources through virtualization technology, while the users may face additional security threats when using virtualization platforms. Co-resident attack is one security problem when an attacker exploits the characteristics of a shared infrastructure and attacks other virtual machines co-located on the same physical machine. In 2017, Abazari et al. proposed a multi-objective response system to against co-resident attacks in cloud environment, considering the co-resident time of virtual machines, and responding to the attacks with the goal of minimum cost and minimum threat. Howeverwe found that the response time of their proposed system was actually too long for real-time applications.In this thesis, we proposed to use machine learning to train the intrusion response system. We use the Ridge Regression algorithm and perform a series of experiments to prove the effect of the proposed model. In the experiments, we also compared our model with that of the Abazari’s. From the experimental results, we showed that our model can obtain a solution with efficiency improvement of 2000x speedup and 87.9% of high accuracy.
