無線微型感測網路上之具防禦力的時間同步協定

Abstract

對無線微型感測網路來說，時間同步是一個重要的前提，因為在眾多的應用程式中，皆需要建立在彼此的時間需一致之上，此類的應用有物體追蹤、速度測量、資料聚合和使用時間槽方式溝通的應用。有一些攻擊會降低現存協定的精確度。此類的攻擊有pulse-delay attack、spoofing attack、replay attack和insider attack。目前已經有很多篇PAPER以安全的角度去設計時間同步協定，但他們並沒有辦法完整抵禦這些攻擊，或是只同步了時間差而沒有同步頻率差。此篇論文介紹了兩種有效率的傳遞方式，一個是使用群播的方式，一個是使用改進過的µTESLA的方式。最後我們提出了一種篩選演算法能濾掉Outlier，我們使用了簡單的評分機制濾掉Extreme outliers然後使用線性回歸線來過濾Mild outliers。在我們的實驗中，我們產生了一些Outliers，然後使用我們的過濾演算法，我們觀察到當過濾比率比Outlier的比率大時，同步完的時間將不會受到Outlier的汙染且是非常準確的。

Time synchronization is considered an important issue for wireless sensor networks (WSNs) where many applications, such as object tracking and data aggregation, rely on the synchronization of the clock of each node. Some types of attacks, such as insider attack and pulse-delay attack, can be used to decrease the precisions of time synchronization protocols. Recently, many time synchronization protocols were designed with security concern. Some of them cannot defend against these attacks, while others cannot synchronize the relative clock skews and offsets simultaneously or cannot support global time synchronization in multi-hop sensor networks. In this paper, we propose two efficient methods to reduce the overhead for propagating the global time. One uses authenticated multicast by inserting multiple MACs into a message. This method is more efficient than the other, but the amount of receivers must be less than eight. The other uses authenticated broadcast based on uTESLA and it doesn’t limit the amount of receivers. The proposed methods can filter the outliers introduced by attacks and keeps the precision about the order of microsecond.