Article Preview
Top1. Introduction
Wireless sensor networks (WSNs) have gained notable more attention from the researchers in recent years. The WSNs have been proposed to be used in a diverse spectrum of applications such as environmental monitoring, military applications, surveillance, healthcare, home automation, control system in the industry, etc. One of the key challenges is ensuring that the WSNs application deployed performs well within the resource constraints of the constituent WSNs nodes. The sensor nodes have limited power, storage, communication and computational capabilities (MicaZ) (TelosB). Data aggregation based on in-network processing is one of the recourse taken in this direction. Data aggregation is based on exploiting the fact that the communication costs are much higher as compared to the computational costs, in general. This is so because, as per a heuristic, a single bit transmission consumes power equal to the execution of 800-1000 instructions (J. Hill, 2000). Hence, instead of communicating multiple packets to the base station from different sensor nodes, sending an aggregated result in one packet could ensure longer life of the deployed WSNs. The aggregation of the data from different sensor nodes is feasible because, often the sensed information contains correlated and redundant data. Overall, data aggregation improves the energy utilization (reduced communication costs), bandwidth utilization (reduced number of packets sent) and the processor utilization (computations are distributed).
However, data aggregation protocols also introduce new security challenges. Since, a single node has now been responsible for summarization of different sensor readings; compromise of such a node renders numerous other sensor readings invalid. With WSNs typically deployed for ubiquitous applications that may involve hostile environments, the likelihood of such attacks viz. node tempering, forged data injection, and false aggregation increases manifold (J. Yick, 2008). Aggregation functions generally used to generate representative value such as min, max, sum and average are not mathematically resilient (D. Wagner, 2004). A single contaminated sample or sub aggregate value allows significant change in the result. For example, consider an aggregator that performs a sum aggregation function with three nodes contributing values 2, 3 and 5. Thus, aggregator calculates the aggregate sum as value 10. Now, if one node is under attack and attacker changes the contribution to 2, 3 and 100, the final aggregated value becomes 105 – that deviates appreciably from the correct result 10. Therefore, it is necessary to devise data aggregation protocols that are resilient against attacks (that tamper with the aggregated value).
In general, the security attributes desired for data aggregation protocols are data confidentiality, privacy, data authentication and data integrity. As we discuss further in section 2, there indeed have been numerous, scattered/isolated attempts to devise secure data aggregation protocols offering
- •
Either confidentiality & privacy (S. Peter, 2010; A. Liu, 2008; Osman Ugus, 2007; Xiaoyan Wang, 2010; Poornima A.S, 2010; Jacques M. Bahi, 2010; Rabindra Bista, 2009; M. Oenen, 2007; Wenbo He, 2007; Giroa J, 2005; Hongjuan Li, 2011; Xinyang Huang, 2007; Hasan Çama, 2006; Suat Ozdemir, 2009; Ajay Jangra, 2013; Taochun W, 2013) or
- •
Data integrity (Julia Albath, 2009; Suat Ozdemir, 2008; Alzaid H, 2008; Wei Zhang, 2006; Bartosz Przydatek, 2007; Lingxuan Hu, 2003; Kui Wu, 2006; Wenliang Du, 2003; Bagaa M, 2012; Shaik M, 2014) or
- •
A combination of these attributes (Suat Ozdemir, 2011; Vimal Kumar, 2010).
However, none of these approaches offers an integrated framework that can support all the minimum functionalities desired out of a secure data aggregation protocol viz. security attributes, topology construction as well as a key management. Prompted by this motivation, we propose a framework for secure data aggregation in this paper that attempts to fill the void.