A Time-Series Pattern Based Noise Generation Strategy for Privacy Protection in Cloud Computing

• Authors: Gaofeng Zhang;Yun Yang;Xiao Liu;Jinjun Chen
• Venue: CCGRID '12 Proceedings of the 2012 12th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (ccgrid 2012)
• Year: 2012

Cloud computing promises an open environment where customers can deploy IT services in a pay-as-you-go fashion while saving huge capital investment in their own IT infrastructure. Due to the openness, various malicious service providers may exist. Such service providers may record service information in a service process from a customer and then collectively deduce the customer's private information. Therefore, from the perspective of cloud computing security, there is a need to take special actions to protect privacy at client sides. Noise obfuscation is an effective approach in this regard by utilising noise data. For instance, it generates and injects noise service requests into real customer service requests so that service providers would not be able to distinguish which requests are real ones if their occurrence probabilities are about the same. However, existing typical noise generation strategies mainly focus on the entire service usage period to achieve about the same final occurrence probabilities of service requests. In fact, such probabilities can fluctuate in a time interval such as three months and may significantly differ than other time intervals. In this case, service providers may still be able to deduce the customers' privacy from a specific time interval although unlikely from the overall period. That is to say, the existing typical noise generation strategies could fail to protect customers' privacy for local time intervals. To address this problem, we develop a novel time-series pattern based noise generation strategy. Firstly, we analyse previous probability fluctuations and propose a group of time-series patterns for predicting future fluctuated probabilities. Then, based on these patterns, we present our strategy by forecasting future occurrence probabilities of real service requests and generating noise requests to reach about the same final probabilities in the next time interval. The simulation evaluation demonstrates that our strategy can cope with these fluctuations to significantly improve the effectiveness of customers' privacy protection.