Modern automotive Cyber-Physical Systems (CPS) are equipped with a variety of vehicular sensors to provide driving convenience (e.g. cruise control systems, navigation systems, and pedestrian detection systems, etc.). However, those systems are vulnerable to the CPS attacks and safety becomes a big recent issue. The malicious sensory data can cause the physical destruction of an actuator, so a robust system against sensor faults is essential for improved safety in Intelligent Transportation Systems (ITS). Therefore, this thesis proposes an anomaly detection mechanism under which performs core functions normally by detecting and filtering out the values from malfunctioning sensors. Herein it is assumed that multiple sensors measure a same physical variable. First, we can detect anomalies among the sensory readings according to the Recursive Least Squares (RLS) algorithm. We show that RLS is more suitable for the real-time operation than Least Mean Square (LMS) algorithm in Adaptive Filter. The RLS algorithm detects the anomalies whose sensor values exceed a threshold. Second, the Iterative Filtering (IF) algorithm determines a reliable average value using the normal incoming sensory readings. The average value is provided to the Proportional-Integral-Derivative (PID) controller, and it (i.e. the current speed) is compared with the reference speed while the vehicle is under the normal cruise control. Without employing the proposed mechanism, according to the experiment, a vehicular model in a simulator had the cruise control system failure and faced a fatal accident. In contrast, the cruise control system under the proposed mechanism operated normally in spite of the injected attack signals. We implemented the robust cruise control system using the combination of RLS and IF algorithms in order to detect anomalies and compute the reliable average value. ⓒ 2016 DGIST
Table Of Contents
I. Introduction 1-- II. System configuration 4-- III. Filtering Mechanism 7-- 3.1. Recursive Least Square (RLS) Filtering Algorithm 8-- 3.1.1. Comparison between RLS and LMS 10-- 3.1.2. Anomaly Detection using RLS 14-- 3.2. Iterative Filtering (IF) Algorithm 16-- 3.2.1. Appropriateness of IF 21-- 3.2.2. Reliable Average Decision using IF 23-- IV. Experiment and Results 26-- 4.1. Experimental Environment 26-- 4.2. Simulation Experiment and Results 29-- V. Conclusion and Future Work 43-- References 45