Sensor attack detection for railway vehicles using topographic information

Attacks on control systems have become an important issue due to recently reported actual cases. As a counter measure, attacks on sensors and their detection have been investigated. The approach of existing work is mostly based on exploiting sensor redundancies. In this work, a sensor attack detection method is suggested using topographic information for railway vehicles. The method detects sensor attack not by using redundancies, but by using topographic information of the railways. Specifically, the method detects sensor attack by comparing predicted disturbance predefined by topographic information or normal operating data, and estimated disturbance calculated by disturbance observer. We verified the method through experiment on a scale train testbed. This method is applicable not only to railway vehicles but also to systems such as vehicle-platooning that repeatedly operate on the same section.