Malicious mining code detection based on ensemble learning in cloud computing environment

Hackers increasingly tend to abuse and nefariously use cloud services by injecting malicious mining code. This malicious code can be spread through infrastructures in the cloud platforms and pose a great threat to users and enterprises. In this study, a method is proposed for detecting malicious mining code in the cloud platforms, which constructs a detection model by fusing the Bagging and Boosting algorithms. By randomly extracting samples and letting models vote together to decide, the variance of model detection can be reduced obviously. Compared with traditional classifiers, the proposed method can obtain higher accuracy and better robustness. The experimental results show that, for the given dataset, the values of AUC and F1-score can reach 0.992 and 0.987 respectively, and the standard deviation of AUC values under different data inputs is only 0.0009.

Automated summary

The study proposes a method for detecting malicious mining code in cloud platforms by fusing Bagging and Boosting algorithms to construct a detection model, reducing the variance of model detection significantly and achieving higher accuracy and robustness compared to traditional classifiers. The experimental results show high values of AUC (0.992) and F1-score (0.987), with a low standard deviation of AUC values under different data inputs (0.0009).