SensorTrust: A resilient trust model for wireless sensing systems

https://doi.org/10.1016/j.pmcj.2010.11.002Get rights and content

Abstract

Wireless sensor networks (WSNs) are prone to failures and malicious attacks. Traditional approaches from encryption and authentication are not sufficient to solve the problems. Trust management of WSNs is bringing new approaches. However, it is still a challenge to establish a trust environment for WSNs. To conquer that challenge, we propose a resilient trust model with a focus on data integrity, SensorTrust, for hierarchical WSNs. SensorTrust integrates past history and recent risk to accurately identify the current trust level. It employs a Gaussian model to rate data integrity in a fine-grained style and a flexible update protocol to adapt to varied context. With acceptable overhead, SensorTrust proves resilient against varied faults and attacks.

Introduction

Wireless sensor networks (WSNs) are prone to failures and malicious attacks. In addition to unstable radio communication, data integrity is also an important issue. Though a secure and robust routing layer may enhance the data delivery ratio, a compromised node itself could report inaccurate or even forged data. Meanwhile, constrained by WSNs resource-starving nature in power, processing and storage, cryptographic methods [1] applied to WSNs are never expected to completely solve the problem.

In contrast to conventional approaches, trust management [2] is becoming a new driving force to solve challenges to ad hoc networks [3], [4], [5], peer-to-peer networks [6], [7], [8], as well as WSNs [9], [10], [11], [12]. Generally, with trust management, each sensor node in the network is assigned a trust value to reflect its trustworthiness according to its past performance. However, as far as WSNs are concerned, there are a few important issues with existing work on trust management. First, most trust research on WSNs focuses on communication behavior, and data integrity is overlooked. Since data collection is the main task of WSNs, the importance of data integrity should never be underestimated. Second, it is difficult to apply those overcomplicated reputation models developed for generic networks to deployed WSNs. Those models may cause much overhead to resource-constrained WSNs; the complexity of the models also causes the difficulty to estimate the impact of parameter choices. Finally, the easy-to-implement trust evaluation scheme that treats new events and past behaviors equally suffers various attacks.

In this paper, we propose a resilient trust model, SensorTrust, for hierarchical WSNs. In this model, the aggregator maintains trust estimations for children nodes. Unlike previous efforts, our current design of SensorTrust mainly focuses on data integrity, though communication robustness can also be incorporated. SensorTrust integrates past history and recent risk in a real-time way that accurately identifies the current trust level. Our model employs a Gaussian model [13] to rate data integrity in a fine-grained style, and a flexible update protocol to adapt to varied context. With acceptable overhead, the SensorTrust model is evaluated with the real world sensor data from Intel Berkeley Lab and Motelab at Harvard University, and compared with other approaches. The results indicate the great advantage of SensorTrust to handle faults and attacks.

The rest of the paper is organized as follows: the detailed mechanism of the SensorTrust model is depicted in Section 3; the efficacy analysis and evaluation of SensorTrust are given in Sections 4 Efficacy of SensorTrust, 5 Evaluation of SensorTrust separately; the related work and our conclusions are presented in Sections 2 Related work, 6 Conclusions. Additionally, a short Appendix describing mathematical details is placed at the end of the paper.

Section snippets

Related work

Trust has been studied in varied contexts for a long time. It started as an important topic in social science. The effects of trust in commerce was analyzed to help build e-commerce systems, such as eBay [14]. Game theory and reinforcement learning are also used to model the reputation of sellers in [15]. Additionally, trust management is applied to online knowledge sharing [16], peer-to-peer systems and ad hoc networks [17], [18], [19], [20], [6].

Regarding WSNs, most current trust research

SensorTrust model

We propose SensorTrust to evaluate the trustworthiness of sensor nodes in hierarchical WSNs. The hierarchical structure has been widely accepted in designing WSNs because it optimizes network performance [25]. In a hierarchical WSN, each node relays data to its associated lower-level aggregator, and those aggregators forward received data to their higher-level parent aggregators, and the forwarding continues until the top layer of the hierarchy, at which point the data will be sent to the base

Efficacy of SensorTrust

In this section, we analyze the efficacy of SensorTrust. As stated above, the SensorTrust model mainly involves two stages: transaction rating and SensorTrust value update.

Regarding transaction rating in terms of data integrity, our SensorTrust model employs the Gaussian model to estimate the integrity of data. Compared to the traditional data outlier detection approaches, our rating is fine-grained, and more precise. Data outlier detection is the main technique to extract data trustworthiness

Evaluation of SensorTrust

In this section, we evaluate SensorTrust with data collected from the Intel Berkeley Research lab [31] and Motelab [32] at Harvard University. To implement SensorTrust onto the aggregator in a WSN, the aggregator maintains a data table to store the data received from its children sensor nodes during the most recent transmission period, and another SensorTrust table to store current SensorTrust values for children nodes. The data table is updated whenever the aggregator receives data, and the

Conclusions

In this paper, we propose a resilient trust model, SensorTrust, for hierarchical WSNs. In this model, the aggregator maintains trust estimations for children nodes. Unlike previous efforts, our current design of SensorTrust mainly focuses on data integrity, though communication robustness can also be incorporated. With this model, past history and recent risk are synthesized in a real-time way that accurately identifies the current trust level. Our model employs the Gaussian model to rate data

References (33)

  • F. Hu et al.

    Security considerations in ad hoc sensor networks

    Ad Hoc Networks

    (2005)
  • Z. Liang et al.

    Analysis of ratings on trust inference in open environments

    Performance Evaluation

    (2008)
  • M. Blaze, J. Feigenbaum, J. Lacy, Decentralized trust management, in: Proceedings of 1996 IEEE Symposium on Security...
  • A. Boukerche, K. El-Khatib, L. Xu, L. Korba, A novel solution for achieving anonymity in wireless ad hoc networks, in:...
  • Z. Liu, A. Joy, R. Thompson, A dynamic trust model for mobile ad hoc networks, in: Proceedings of the 10th IEEE...
  • Q. He, D. Wu, P. Khosla, Sori: a secure and objective reputation-based incentive scheme for ad hoc networks, in:...
  • S. Kamvar, M. Schlosser, H. Garcia-Molina, The eigentrust algorithm for reputation management in P2P networks, in:...
  • Y. Wang, J. Vassileva, Trust and reputation model in peer-to-peer networks, in: Proceedings of the 3rd International...
  • Z. Liang et al.

    Pet: a personalized trust model with reputation and risk evaluation for P2P resource sharing

  • W. Cheng et al.

    A trust-based routing framework in energy-constrained wireless sensor networks

  • N. Lewis et al.

    Using trust for key distribution and route selection in wireless sensor networks

  • S. Tanachaiwiwat, P. Dave, R. Bhindwale, A. Helmy, Secure locations: routing on trust and isolating compromised sensors...
  • H. Wang et al.

    Trust routing protocol framework based on behavior assessment for wireless sensor networks

  • O. Kallenberg

    Foundations of Modern Probability

    (2002)
  • P. Resnick, R. Zeckhauser, Trust among strangers in Internet transactions: empirical analysis of ebays reputation...
  • T. Tran et al.

    A reputation coriented reinforcement learning strategy for agents in electronic marketplaces

    Computational Intelligence

    (2002)
  • Cited by (0)

    A poster abstract of this work was presented at the 7th ACM Conference on Embedded Networked Sensor Systems (SenSys 2009). This work is in part supported by the National Science Foundation grant CNS-0721456.

    View full text