User-Level Side Channel Attack on Workflow System in Data-Center

Though current workflow systems, known as the major task managing tool in data-center, use UNIX-based user/group authorization mechanisms to defend unauthorized operations, the information in workflow systems, such as DAG (Directed Acyclic Graph), could be easily exposed to any user-level malicious monitor. In this paper, we propose a type of side channel attack to data-center workflow information. By this attack, the DAG of workflow systems can be stolen through monitoring basic system characteristics, such as CPU utilization, memory accessing, disk I/O, etc, which can successfully circumvents UNIX privilege checking. Also, we present a software-diversity based measurecounter to mitigate the attack.