Attacker Detection Based on Multi-Level Tenant Classification

Co-resident attacks in the cloud environment have always been a concern. Different from other network attacks, it is a cache-based side-channel attack. Many studies have shown that it can be co-located with the target user on a server, and then analyze some cached data to steal or tamper with user data, thereby launching an attack. Therefore, how to detect and judge malicious users is an urgent issue. Researchers have proposed a variety of detection and defense methods. This paper proposes a detection model to detect and classify malicious users according to abnormal hardware events. The Deep Belief Network (DBN) is responsible for extracting features, while the Support Vector Machine (SVM) handles the classification. Extracting features improves classification efficiency while maintaining high accuracy. According to the experimental results, the proposed method can classify malicious users, and the accuracy rate reaches 99.49 % and the classification speed demonstrates performance that is on par with existing methods.