A Transparent Learning Approach for Attack Prediction Based on User Behavior Analysis

摘要

User behavior can be used to determine vulnerable user actions and predict potential attacks. To our knowledge, much work has focused on finding vulnerable operations and disregarded reasoning/-explanations of its results. This paper proposes a transparent learning approach for user behavior analysis to address this issue. A user rating system is proposed to determine a security level of each user from several aspects, augmented with explanations of potential attacks based on his/her vulnerable user actions. This user rating model can be constructed by a semi-supervised learning classifier, and a rule mining algorithm can be applied to find hidden patterns and relations between user operations and potential attacks. With this approach, an organization can be aware of its weakness, and can better prepare for proactive attack defense or reactive responses.