Post Abstract: Role-based Deceptive Detection and Filtering in WSNs Shinan Wang Wayne State University shinan@wayne.edu Kewei Sha Oklahoma City University ksha@okcu.edu Weisong Shi Wayne State University weisong@wayne.edu 1. PROBLEM STATEMENT With more and more real applications of WSNs have been deployed, which are in charge of either monitor- ing parameters or event detection, we envision that the success of the WSNs is decided by the quality of the collected data [2]. Further more, the quality of the col- lected data is mainly affected by the deceptive data, which includes redundant data, very similar repeated readings that provide less information, and false data, wrong readings resulted from unreliable components or wireless communication and malicious attacks. On one hand, redundant data should be filtered because less information is provided but considerable resources are wasted. On the other hand, false data should also be filtered to improve data accuracy. Hence, the major concern of improving the data qual- ity is to detect and filter deceptive data. However, be- cause of limited resources, it is a big challenge to imple- ment deceptive data detection, which is further compli- cated when system has high dynamics. Several schemes have been developed to solve this issue [1, 3], but lots of them concerns rather on the systems than the data itself. Moreover, few of them targets deceptive data de- tection in high dynamic systems. Thus, we are targeting to detect and filter the deceptive data in high-dynamic event-driven WSNs from the data itself point of view by proposing a Role-based Deceptive Detection and Filter- ing (RD 4 ) mechanism in this paper. The detail of our approach is listed as follows 2. OUR APPROACH First, each sensor picks up a role from the role set Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Copyright 2009 ACM 978-1-60558-371-6/09/04 ...$5.00. based on the specific features of the sensor, such as storage size, computation ability, communication ability and trustable level. For each event it sensed or received, the sensor issues a confidence score to the event, which is denoted as csr(E,T ) ij , indicating the truth level of this event, where E specifies the event, and i and j are the identity of the role and the identity of the sensor, while T means the score will be valid for T time slots. Moreover, different roles have different maximum confi- dent scores that can be issued toward an event. In RD 4 , the confidential score defined above is calcu- lated based on the accumulated signal strength during a certain time period ([0,T 0 ]), depicted as ASS (E,T 0 ) ij , of the corresponding event E at sensor j with role i. Here, the signal strength of an event E, denoted as SS (E) ij , can be defined as the amount of changes of a monitoring physical parameter within a unit time pe- riod. For example, if we try to detect an event of sud- den changes in temperature at a computing node in a high performance computing system, the signal strength will be the amount of temperature changing within each minute. Thus, if the function to specify the changing rate of a monitoring physical parameter is p(t), we can define SS (E) ij as SS (E) i,j = p(t)dt (1) Based on the defined SS (E) ij , the accumulated signal strength can be defined as ASS (E,T 0 ) ij =  T0 0 SS (E) ij =  T0 0 p(t)dt (2) Having ( 2), we design a function f that maps the accumulated signal strength to a confidential score, to be specific, csr(E,T ) ij = f (ASS (E,T 0 ) ij ). When an event is detected at a sensor, the sensor will set up a timer T , also used as the first lifetime period of the event, to the detected event. Then the sensor will try to confirm whether it is a real event before the event expires. The decision is made based on the confidence score, which can come from two sources. One is the Copyright is held by author/owner IPSN’09, April 13–16, 2009, San Francisco, California, USA. ACM 978-1-60558-371-6 387