Research and Practice on Detection of Abnormal Campus Accounts Based on User Sign-in Logs of Business Systems
Wenfeng Zhong, Ying Zhang, HuaiChu Chen · 2022
In recent years, information security has been getting more concerned, and timely detection of abnormal campus accounts has great significance to the information security of colleges and universities. This paper proposes an abnormal campus accounts detection method using a machine learning model to detect abnormal campus accounts that have suspicious activities, and uses the quantity and the percentage of user sign-in logs of different campus business systems to detect accounts that have unusual access behaviors. This paper uses user sign-in logs of campus business systems for 9 consecutive months and the known abnormal campus accounts as data sources, and generates samples based on the proposed method to train the machine learning models, then verifies the detection effect of the trained machine learning models. The results show that the proposed method achieves a detection rate of up to 75% with a false-positive rate of 0.26% in identifying abnormal campus accounts.