Securing Ultrasonic Sensors Against Signal Injection Attacks Based on a Mathematical Model

• Main Authors: Sunwoo Lee, Wonsuk Choi, Dong Hoon Lee
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Acoustic sensors; mathematical model; physical layer; security; sensor systems; signal analysis
• Online Access: https://ieeexplore.ieee.org/document/8786810/

Ultrasonic sensors are used across a variety of industrial areas, as they measure distance simply via calculating the difference of time between transmission and reception of signals. However, recently, attackers have been targeting ultrasonic sensors to inflict intentional malfunction regarding distance measurement to an obstacle and some researchers have actually demonstrated such attacks using an actual vehicle, the Tesla Model S. In addition, these malicious signal injections into the ultrasonic sensors do not require sophisticated equipment, making it simple to stage a jamming or spoofing attack. This means that, in practice, signal injection attacks on ultrasonic sensors are possible, and, as such, detection of these attacks is crucial. Owing to this possibility, several methods to secure sensors have been proposed. However, these approaches cannot be applied directly to ultrasonic sensors or can operate only in specific, controlled environments. Here we propose an ultrasonic-sensor-specific method that is capable of detecting signal injection attacks without modifying the structural design of the existing environment in which an ultrasonic sensor is applied. Our method is proven to work with a single ultrasonic sensor and also operates when a sensor and/or target are in motion. In addition, we present a mathematical model for detecting maliciously injected signals based on the properties of the sensor system and the relation between signals transmitted and received from the sensor. We evaluate our method using a commercial ultrasonic sensor to demonstrate its efficacy to detect jamming attacks and spoofing attacks under real conditions. Finally, our method does not require additional devices or substantial resources.