Speculative reordering for a latency-optimized privacy protection in complex event processing

Abstract

With increasing number of applications in Internet of Things (IoT), Complex Event Processing (CEP) has already become one of the state-of-the-art technologies recently. In CEP, privacy needs to be considered carefully because events with user’s sensitive information may be exposed to outside world. However, most privacy issues in CEP mainly focus on attribute-based events without considering pattern-based events. There are two important works for pattern-based privacy in CEP: suppression and re-ordering. The former suppresses events belonging to private patterns while the later tends to reorder them. The re-ordering mechanism shows better performance in terms of QoS, but the latency would be long when the size of window increases. Also, the re-ordering strategy is performed only at the end of the windows. In this thesis, we extend the Re-ordering strategy by using speculation based on Markov chains, so we start speculating whether the private pattern occurs in current window before the end of the window. If the private pattern is predicted to occur, we then already re-order events that are part of private patterns. Additionally, the top-k preserving algorithm is introduced for preserving public patterns. Our evaluation results show that we maintain nearly 80 % utility when compared to the normal re-ordering strategy. From our experiments, it is seen that we can eliminate the time taken for re-ordering completely if the window size is greater than 3 ms.