A Two-Stage Classification Model on Detecting Cache-Based Side-Channel Attacks

Cache-based side-channel attacks in cloud environments have attracted widespread attention. Many studies have successfully proved that sensitive data can be stolen from co-resident virtual machines by launching these attacks. Relatively various detection and defense methods are also proposed. To address such problems, this paper proposes a novel two-stage classification model that combines random forest and XGBoost, which can classify the samples collected from one process into attack samples, noise samples and normal samples. In the experiments, the accuracy of attack samples and normal samples produced results of 98.89% and 99.83%, respectively, demonstrating that our model is not inferior compared to the current methods in terms of classification performance and outperforms them in terms of implementation difficulty and classification speed.