過濾錯誤讀數以增進無線感測器網路資料正確性

Abstract

在這篇論文中，我們提出了一個有效率的方法來偵測感測器網路上的錯誤讀數，在不會遺漏因環境改變所造成的讀數變化的情況下。更明確的說，首先我們發掘出感測器網路中，感測器彼此在空間上的相關性以及讀數上行為的異同，導出感測器彼此之間的相似程度。根據這些相似度的關係，再利用馬可夫鏈(Markov chain)來將所有的感測器排序並且計算出相對應的排序值SensorRank。依據SensorRank，我們提出了一個很有效率的演算法(TrustVoting)解決錯誤讀數偵測的問題。最後，我們也提出了各種的實驗來證明我們的方法以及SensorRank的優點。

In this paper, we propose an effective filtering readings detection algorithm to judiciously determine faulty readings without missing interesting events in sensor networks. Explicitly, by exploring spatial correlation and reading behaviors of sensors, we first derive similarity relationships (referred to as trust relation) among sensors. Then, we model the relationships of sensors as a Markov chain and thus develop SensorRank, a mechanism for rating sensors in terms of similarity relationships. In light of SensorRank, we propose a filtering algorithm to effectively detect and filter faulty readings. Performance studies are conducted and simulation results show that our proposed algorithm outperforms others, showing the advantage of SensorRank.