Determination of Favorable Factors for Cloud IP Recognition Technology

Identifying cloud IP usage scenarios is critical for cybersecurity applications, yet existing machine learning methods rely heavily on numerous features, resulting in high complexity and low interpretability. To address these issues, this paper proposes an approach to identify cloud IPs from the perspective of network attributes. We employ data mining and crowdsourced collection strategies to gather IP addresses from various usage scenarios, which including cloud IPs and non-cloud IPs. On this basis, we establish a cloud IP identification feature set that includes attributes such as Autonomous System Number (ASN) and organization information. By analyzing the differences in the properties of different IP usage scenarios in the detection results, we can find out the factors that are conducive to cloud IP identification. Experimental evaluation demonstrates that the proposed method achieves a high identification accuracy of 96.67%, surpassing the performance of traditional machine learning models such as CNN, MLP, XGBoost, KNN, SVM, and Decision Tree, whose accuracies range between 81% and 92%. Furthermore, this study reveals that latency and port information exhibit insufficient discrimination power for distinguishing cloud IP from non-cloud IP scenarios, highlighting ASN as a simpler, more interpretable, and resource-efficient criterion. To facilitate reproducible research, datasets and codes are publicly released.