Translator Disclaimer
23 May 2011 Sensor trustworthiness in uncertain time varying stochastic environments
Author Affiliations +
Persistent surveillance applications require unattended sensors deployed in remote regions to track and monitor some physical stimulant of interest that can be modeled as output of time varying stochastic process. However, the accuracy or the trustworthiness of the information received through a remote and unattended sensor and sensor network cannot be readily assumed, since sensors may get disabled, corrupted, or even compromised, resulting in unreliable information. The aim of this paper is to develop information theory based metric to determine sensor trustworthiness from the sensor data in an uncertain and time varying stochastic environment. In this paper we show an information theory based determination of sensor data trustworthiness using an adaptive stochastic reference sensor model that tracks the sensor performance for the time varying physical feature, and provides a baseline model that is used to compare and analyze the observed sensor output. We present an approach in which relative entropy is used for reference model adaptation and determination of divergence of the sensor signal from the estimated reference baseline. We show that that KL-divergence is a useful metric that can be successfully used in determination of sensor failures or sensor malice of various types.
© (2011) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Ajay Verma, Ronald Fernandes, and Kalyan Vadakkeveedu "Sensor trustworthiness in uncertain time varying stochastic environments", Proc. SPIE 8047, Ground/Air Multisensor Interoperability, Integration, and Networking for Persistent ISR II, 80470O (23 May 2011);

Back to Top