WSN Data Confidence Attribution Using Predictors
Roberto Milton Scheffel, Antnio A. Frhlich · 2018
Wireless Sensor Networks are becoming the communication base for many Cyber-Physical Systems, which rely on sensed data to make sensitive decisions. Faulty data can lead such systems to unpredictable behavior. It can result from sensor's hardware failures and also from the intentional interference of an intruder. Gateways connecting such CPS with the Internet usually run conventional operating systems and communication protocols and therefore are natural candidates to be overtaken by intruders. This work proposes the use of a confidence attribution scheme, based on lightweight predictors, capable of detecting wrong sensor data caused either by sensor failures or by malicious data forging. Simulations show that its possible to achieve high detection rates with a very low communication overhead.