CS-1-SVM: Improved One-class SVM for Detecting API Abuse on Open Network Service

Abstract

Open network service is threatened by API abusers such as spammers, phishes, compromised users, etc., be- cause of their open API for any user and third-party developers. In order to preserve the service resource and security, we proposed an approach called CS-1-SVM based on cosine similarity and 1-SVM to detect anomalous accounts who abused API in open network service. Two of the key processes of the method are account modeling and classifier solving. In ac- count modeling, we vectorized every sample user by extracting the dynamic features and calculating the cosine similarity between static features. In classifier solving, we improved 1-SVM in regularization parameter optimization efficiency with cosine similarity too. Based on the proposed method, we developed an experiment to demonstrate that CS-1-SVM has the ability to detect both malicious and compromised account and simplify the process of parameter optimization without reducing the accuracy of 1-SVM.